Relativistic Many-Body Theory: A New Field-Theoretical Approach

Springer Series on Atomic, Optical, and Plasma Physics Volume 63 Series Editors W.E. Baylis Uwe Becker Philip George B...

Author: Ingvar Lindgren

49 downloads 1187 Views 2MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Springer Series on Atomic, Optical, and Plasma Physics Volume 63

Series Editors W.E. Baylis Uwe Becker Philip George Burke Robert N. Compton M.R. Flannery Charles J. Joachain Brian R. Judd Kate Kirby Peter Lambropoulos Gerd Leuchs Pierre Meystre

For further volumes: http://www.springer.com/series/411

Ingvar Lindgren

Relativistic Many-Body Theory A New Field-Theoretical Approach With 97 Figures

ABC

Ingvar Lindgren Department of Physics University of Gothenburg SE-412 96 Göteborg Sweden [email protected]

ISSN 1615-5653 ISBN 978-1-4419-8308-4 e-ISBN 978-1-4419-8309-1 DOI 10.1007/978-1-4419-8309-1 Springer New York Dordrecht Heidelberg London Library of Congress Control Number: 2011921721 c Springer Science+Business Media, LLC 2011 All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the publisher (Springer Science+Business Media, LLC, 233 Spring Street, New York, NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed is forbidden. The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights. Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

To Eva

Preface

It is now almost 30 years since the first edition of my book together with John Morrison, Atomic Many-Body Theory [6], appeared, and the second edition appeared some years later. It has been out of print for quite some time, but fortunately has recently been made available again through a reprint by Springer Verlag. During the time that has followed, there has been a tremendous development in the treatment of many-body systems, conceptually as well as computationally. Particularly the relativistic treatment has expanded considerably, a treatment that has been extensively reviewed recently by Ian Grant in the book Relativistic Quantum Theory of Atoms and Molecules [2]. Also, the treatment of quantum-electrodynamical (QED) effects in atomic systems has developed considerably in the last few decades, and several review articles have appeared in the field [7, 11, 13] besides the book by Labzowsky et al., Relativistic Effects in Spectra of Atomic Systems [5]. An impressive development has taken place in the field of many-electron systems by means of various coupled-cluster approaches, with applications particularly on molecular systems. The development during the last 50 years has been summarized ˇ arsky, Paldus, in the book Recent Progress in Coupled Cluster Methods, edited by C´ and Pittner [14]. The present book is aimed at combining atomic many-body theory with quantumelectrodynamics, which is a long-sought goal in quantum physics. The main problem in this effort has been that the methods for QED calculations, such as the S-matrix formulation, and the methods for many-body perturbation theory (MBPT) have completely different structures. With the development of the new method for QED calculations, the covariant evolution operator formalism by the Gothenburg atomic theory group [7], the situation has changed, and quite new possibilities appeared to formulate a unified theory. The new formalism is based on field theory, and in its full extent the unification process represents a formidable problem, and we can in this book describe only how some steps toward this goal can be taken. This book is largely based upon the previous book on Atomic Many Body Theory [6], and it is assumed that the reader has absorbed most of that book, particularly Part II. In addition, the reader is expected to have basic knowledge in quantum field theory that is explained in

vii

viii

Preface

books such as Quantum Theory of Many-Particle Systems by Fetter and Walecka [1] (mainly parts I and II), An introduction to Quantum Field Theory by Peskin and Schroeder [12], and Quantum Field Theory by Mandl and Shaw [10]. The material of this book is largely based upon lecture notes and recent publications by the Gothenburg Atomic-Theory Group [3, 4, 7–9], and I want to express my sincere gratitude particularly to my previous coauthor John Morrison and to my present coworkers, Sten Salomonson and Daniel Hedendahl, as well as to the previous collaborators Ann-Marie Pendrill, Jean-Louis Heully, Eva Lindroth, Björn ˚ en, Hans Persson, Per Sunnergren, Martin Gustavsson, and H˚akan Warston for As´ valuable collaboration. In addition, I want to thank the late pioneers of the field, Per-Olov Löwdin, who taught me the foundations of perturbation theory some 40 years ago, and Hugh Kelly, who introduced the diagrammatic representation into atomic physics – two corner stones of the later developments. Furthermore, I have benefitted greatly from communications with many other national and international colleagues and friends (in alphabetic order), Rod Bartlett, Erkki Brändas, Gordon Drake, Ephraim Eliav, Stephen Fritzsche, Gerald Gabrielse, Walter Greiner, Paul Indelicato, Karol Jankowski, Jürgen Kluge, Leonti Labzowsky, Peter Mohr, Debashis Mukherjee, Marcel Nooijen, Joe Paldus, Vladimir Shabaev, Thomas Stöhlker, Gerhard Soff , Joe Sucher, Peter Surjan, and many others. The outline of the book is the following. The main text is divided into three parts. Part I gives some basic formalism and the basic many-body theory that will serve as a foundation for the following. In Part II, three numerical procedures for calculation of QED effects on bound electronic states are described, the S-matrix formulation, the Green’s-function, and the covariant-evolution-operator methods. A procedure toward combining QED with MBPT is developed in Part III. Part IV contains a number of Appendices, where basic concepts are summarized. Certain sections of the text that can be omitted at first reading are marked with an asterisk (*).

References 1. Fetter, A.L., Walecka, J.D.: The Quantum Mechanics of Many-Body Systems. McGraw-Hill, N.Y. (1971) 2. Grant, I.P.: Relativistic Quantum Theory of Atoms and Molecules. Springer, Heidelberg (2007) 3. Hedendahl, D.: Towards a Relativistic Covariant Many-Body Perturbation Theory. Ph.D. thesis, University of Gothenburg, Gothenburg, Sweden (2010) 4. Hedendahl, D., Salomonson, S., Lindgren, I.: . Phys. Rev. p. (to be published) (2011) 5. Labzowski, L.N., Klimchitskaya, G., Dmitriev, Y.: Relativistic Effects in Spactra of Atomic Systems. IOP Publ., Bristol (1993) 6. Lindgren, I., Morrison, J.: Atomic Many-Body Theory. Second edition, Springer-Verlag, Berlin (1986, reprinted 2009) ˚ en, B.: The covariant-evolution-operator method in bound7. Lindgren, I., Salomonson, S., As´ state QED. Physics Reports 389, 161–261 (2004) 8. Lindgren, I., Salomonson, S., Hedendahl, D.: Many-body-QED perturbation theory: Connection to the two-electron Bethe-Salpeter equation. Einstein centennial review paper. Can. J. Phys. 83, 183–218 (2005)

Preface

ix

9. Lindgren, I., Salomonson, S., Hedendahl, D.: Many-body procedure for energy-dependent perturbation: Merging many-body perturbation theory with QED. Phys. Rev. A 73, 062,502 (2006) 10. Mandl, F., Shaw, G.: Quantum Field Theory. John Wiley and Sons, New York (1986) 11. Mohr, P.J., Plunien, G., Soff, G.: QED corrections in heavy atoms. Physics Reports 293, 227–372 (1998) 12. Peskin, M.E., Schroeder, D.V.: An introduction to Quantun Field Theory. Addison-Wesley Publ. Co., Reading, Mass. (1995) 13. Shabaev, V.M.: Two-times Green’s function method in quantum electrodynamics of high-Z few-electron atoms. Physics Reports 356, 119–228 (2002) ˇ arsky, P., Paldus, J., Pittner, J. (eds.): Recent Progress in Coupled Cluster Methods: Theory 14. C´ and Applications. Springer (2009)

Gothenburg, Sweden

Ingvar Lindgren

Contents

1

Introduction .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 1.1 Standard Many-Body Perturbation Theory . . . . . . . . . .. . . . . . . . . . . . . . . . . 1.2 Quantum-Electrodynamics.. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 1.3 Bethe–Salpeter Equation .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 1.4 Helium Atom: Analytical Approach . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 1.5 Field-Theoretical Approach to Many-Body Perturbation Theory.. . . References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .

1 1 2 3 4 5 7

Part I Basics: Standard Many-Body Perturbation Theory 2

Time-Independent Formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 2.1 First Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 2.1.1 De Broglie’s Relations . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 2.1.2 The Schrödinger Equation . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 2.2 Second Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 2.2.1 Schrödinger Equation in Second Quantization .. . . . . . . . . . . . . 2.2.2 Particle–Hole Formalism: Normal Order and Contraction.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 2.2.3 Wick’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 2.3 Time-Independent Many-Body Perturbation Theory .. . . . . . . . . . . . . . . . 2.3.1 Bloch Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 2.3.2 Partitioning of the Hamiltonian . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 2.4 Graphical Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 2.4.1 Goldstone Diagrams .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 2.4.2 Linked-Diagram Expansion .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 2.5 All-Order Methods: Coupled-Cluster Approach .. . . .. . . . . . . . . . . . . . . . . 2.5.1 Pair Correlation .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 2.5.2 Exponential Ansatz .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 2.5.3 Various Models for Coupled-Cluster Calculations: Intruder-State Problem .. . . . . .. . . . . . . . . . . . . . . . . 2.6 Relativistic MBPT: No-Virtual-Pair Approximation . . . . . . . . . . . . . . . . . 2.6.1 QED Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .

13 13 13 14 16 16 18 19 20 20 21 25 25 28 30 30 33 36 37 39

xi

xii

Contents

2.7

Some Numerical Results of Standard MBPT and CC Calculations, Applied to Atoms . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 40 References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 43

3

Time-Dependent Formalism .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 3.1 Evolution Operator .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 3.2 Adiabatic Damping: Gell-Mann–Low Theorem .. . . .. . . . . . . . . . . . . . . . . 3.2.1 Gell-Mann–Low Theorem . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 3.3 Extended Model Space: The Generalized Gell-Mann–Low Relation .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .

47 47 51 52 52 56

Part II Quantum-Electrodynamics: One-Photon and Two-Photon Exchange 4

S-Matrix . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.1 Definition of the S-Matrix: Feynman Diagrams . . . . .. . . . . . . . . . . . . . . . . 4.2 Electron Propagator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.3 Photon Propagator .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.3.1 Feynman Gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.3.2 Coulomb Gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.4 Single-Photon Exchange .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.4.1 Covariant Gauge .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.4.2 Noncovariant Coulomb Gauge .. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.4.3 Single-Particle Potential .. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.5 Two-Photon Exchange . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.5.1 Two-Photon Ladder.. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.5.2 Two-Photon Cross . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.6 QED Corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.6.1 Bound-Electron Self-Energy .. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.6.2 Vertex Correction .. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.6.3 Vacuum Polarization.. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.6.4 Photon Self-Energy .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.7 Feynman Diagrams for the S-Matrix: Feynman Amplitude .. . . . . . . . . 4.7.1 Feynman Diagrams .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4.7.2 Feynman Amplitude .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .

59 60 61 65 66 68 69 69 72 74 75 75 78 79 79 82 84 87 87 87 88 89

5

Green’s Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 5.1 Classical Green’s Function.. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 5.2 Field-Theoretical Green’s Function: Closed-Shell Case . . . . . . . . . . . . . 5.2.1 Definition of the Field-Theoretical Green’s Function.. . . . . . 5.2.2 Single-Photon Exchange . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 5.2.3 Fourier Transform of the Green’s Function . . . . . . . . . . . . . . . . .

91 91 92 92 95 96

Contents

xiii

5.3

Graphical Representation of the Green’s Function . .. . . . . . . . . . . . . . . . .100 5.3.1 Single-Particle Green’s Function . . . . . . . . . . .. . . . . . . . . . . . . . . . .100 5.3.2 Many-Particle Green’s Function . . . . . . . . . . . .. . . . . . . . . . . . . . . . .105 5.3.3 Self-Energy: Dyson Equation .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .108 5.3.4 Numerical Illustration . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .110 5.4 Field-Theoretical Green’s Function: Open-Shell Case . . . . . . . . . . . . . . .110 5.4.1 Definition of the Open-Shell Green’s Function . . . . . . . . . . . . .110 5.4.2 Two-Times Green’s Function of Shabaev . .. . . . . . . . . . . . . . . . .111 5.4.3 Single-Photon Exchange . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .114 References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .117 6

Covariant Evolution Operator and Green’s Operator . .. . . . . . . . . . . . . . . . .119 6.1 Definition of the Covariant Evolution Operator . . . . .. . . . . . . . . . . . . . . . .119 6.2 Single-Photon Exchange in the Covariant-Evolution-Operator Formalism . . . . . . . . . . .. . . . . . . . . . . . . . . . .122 6.2.1 Single-Photon Ladder . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .126 6.3 Multiphoton Exchange .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .127 6.3.1 General .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .127 6.3.2 Irreducible Two-Photon Exchange.. . . . . . . . .. . . . . . . . . . . . . . . . .128 6.3.3 Potential with Radiative Parts . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .130 6.4 Relativistic Form of the Gell-Mann–Low Theorem .. . . . . . . . . . . . . . . . .131 6.5 Field-Theoretical Many-Body Hamiltonian.. . . . . . . . .. . . . . . . . . . . . . . . . .132 6.6 Green’s Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .134 6.6.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .134 6.6.2 Relation Between the Green’s Operator and Many-Body Perturbation Procedures . .. . . . . . . . . . . . . . . . .136 6.7 Model-Space Contribution .. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .138 6.7.1 Lowest Orders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .139 6.7.2 All Orders .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .142 6.8 Bloch Equation for Green’s Operator . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .146 6.9 Time Dependence of the Green’s Operator. Connection to the Bethe–Salpeter Equation .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .151 6.9.1 Single-Reference Model Space . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .151 6.9.2 Multireference Model Space . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .154 References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .156

7

Numerical Illustrations to Part II. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .157 7.1 S-Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .157 7.1.1 Electron Self-Energy of Hydrogen-Like Ions . . . . . . . . . . . . . . .157 7.1.2 Lamb Shift of Hydrogen-Like Uranium .. . .. . . . . . . . . . . . . . . . .158 7.1.3 Lamb Shift of Lithium-Like Uranium . . . . . .. . . . . . . . . . . . . . . . .159 7.1.4 Two-Photon Nonradiative Exchange in Helium-Like Ions . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .160 7.1.5 Electron Correlation and QED Calculations on Ground States of Helium-Like Ions . . . . .. . . . . . . . . . . . . . . . .163

xiv

Contents

7.1.6

g-Factor of Hydrogen-Like Ions: Mass of the Free Electron . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .164 7.2 Green’s-Function and Covariant-Evolution-Operator Methods .. . . . .165 7.2.1 Fine-Structure of Helium-Like Ions . . . . . . . .. . . . . . . . . . . . . . . . .165 7.2.2 Energy Calculations of 1s2s Levels of Helium-Like Ions .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .167 References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .168 Part III

Quantum-Electrodynamics Beyond Two-Photon Exchange: Field-Theoretical Approach to Many-Body Perturbation Theory

8

Covariant Evolution Combined with Electron Correlation.. . . . . . . . . . . . .173 8.1 General Single-Photon Exchange . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .173 8.1.1 Transverse Part.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .174 8.1.2 Coulomb Interaction .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .181 8.2 General QED Potential .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .181 8.2.1 Single Photon with Crossed Coulomb Interaction . . . . . . . . . .181 8.2.2 Electron Self-Energy and Vertex Correction . . . . . . . . . . . . . . . .186 8.2.3 Vertex Correction with Further Coulomb Iterations . . . . . . . .192 8.2.4 General Two-Body Potential . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .193 8.3 Unification of the MBPT and QED Procedures: Connection to Bethe–Salpeter Equation . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .193 8.3.1 MBPT–QED Procedure . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .193 8.4 Coupled-Cluster-QED Expansion . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .196 References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .198

9

The Bethe–Salpeter Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .199 9.1 The Original Derivations by the Bethe–Salpeter Equation .. . . . . . . . . .199 9.1.1 Derivation by Salpeter and Bethe . . . . . . . . . . .. . . . . . . . . . . . . . . . .199 9.1.2 Derivation by Gell-Mann and Low . . . . . . . . .. . . . . . . . . . . . . . . . .202 9.1.3 Analysis of the Derivations of the Bethe–Salpeter Equation . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .203 9.2 Quasi-Potential and Effective-Potential Approximations: Single-Reference Case. . . . . . . . . . . . .. . . . . . . . . . . . . . . . .205 9.3 Bethe–Salpeter–Bloch Equation: Multireference Case . . . . . . . . . . . . . . .206 9.4 Problems with the Bethe–Salpeter Equation .. . . . . . . .. . . . . . . . . . . . . . . . .208 References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .209

10 Implementation of the MBPT–QED Procedure with Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .211 10.1 The Fock-Space Bloch Equation . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .211 10.2 Single-Photon Potential in Coulomb Gauge: No Virtual Pairs . . . . . . .213 10.3 Single-Photon Exchange: Virtual Pairs. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .216 10.3.1 Illustration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .216

Contents

xv

10.3.2 Full Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .219 10.3.3 Higher Orders .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .221 10.4 Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .221 10.4.1 Two-Photon Exchange . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .221 10.4.2 Beyond Two Photons .. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .221 10.4.3 Outlook.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .224 References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .224 11 Analytical Treatment of the Bethe–Salpeter Equation . .. . . . . . . . . . . . . . . . .225 11.1 Helium Fine Structure.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .225 11.2 The Approach of Sucher . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .226 11.3 Perturbation Expansion of the BS Equation . . . . . . . . .. . . . . . . . . . . . . . . . .231 11.4 Diagrammatic Representation . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .233 11.5 Comparison with the Numerical Approach . . . . . . . . . .. . . . . . . . . . . . . . . . .235 References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .235 12 Regularization and Renormalization .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .237 12.1 The Free-Electron QED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .237 12.1.1 The Free-Electron Propagator.. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .237 12.1.2 The Free-Electron Self-Energy . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .239 12.1.3 The Free-Electron Vertex Correction . . . . . . .. . . . . . . . . . . . . . . . .240 12.2 Renormalization Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .242 12.2.1 Mass Renormalization .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .242 12.2.2 Charge Renormalization.. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .244 12.3 Bound-State Renormalization: Cutoff Procedures .. .. . . . . . . . . . . . . . . . .248 12.3.1 Mass Renormalization .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .248 12.3.2 Evaluation of the Mass Term . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .249 12.3.3 Bethe’s Nonrelativistic Treatment . . . . . . . . . .. . . . . . . . . . . . . . . . .251 12.3.4 Brown–Langer–Schaefer Regularization . . .. . . . . . . . . . . . . . . . .252 12.3.5 Partial-Wave Regularization . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .255 12.4 Dimensional Regularization in Feynman Gauge.. . . .. . . . . . . . . . . . . . . . .257 12.4.1 Evaluation of the Renormalized Free-Electron Self-Energy in Feynman Gauge .. . . . . . . . . . . .. . . . . . . . . . . . . . . . .258 12.4.2 Free-Electron Vertex Correction in Feynman Gauge . . . . . . .261 12.5 Dimensional Regularization in Coulomb Gauge . . . .. . . . . . . . . . . . . . . . .263 12.5.1 Free-Electron Self-Energy in the Coulomb Gauge.. . . . . . . . .263 12.6 Direct Numerical Regularization of the Bound-State Self-Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .267 12.6.1 Feynman Gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .268 12.6.2 Coulomb Gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .268 References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .269 13 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .271

xvi

Contents

Appendices A

Notations and Definitions .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .273 A.1 Four-Component Vector Notations . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .273 A.2 Vector Spaces .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .275 A.2.1 Notations .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .275 A.2.2 Basic Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .275 A.2.3 Special Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .277 A.3 Special Functions.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .277 A.3.1 Dirac Delta Function . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .277 A.3.2 Integrals over Functions . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .279 A.3.3 The Heaviside Step Function . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .282 References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .282

B

Second Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .283 B.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .283 B.2 Heisenberg and Interaction Pictures .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .286 References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .287

C

Representations of States and Operators . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .289 C.1 Vector Representation of States. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .289 C.2 Matrix Representation of Operators .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .291 C.3 Coordinate Representations .. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .292 C.3.1 Representation of Vectors . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .292 C.3.2 Closure Property .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .292 C.3.3 Representation of Operators . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .293

D

Dirac Equation and the Momentum Representation . . . .. . . . . . . . . . . . . . . . .295 D.1 Dirac Equation.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .295 D.1.1 Free Particles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .295 D.1.2 Dirac Equation in an Electromagnetic Field . . . . . . . . . . . . . . . .300 D.2 Momentum Representation . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .300 D.2.1 Representation of States . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .300 D.2.2 Representation of Operators . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .301 D.2.3 Closure Property for Momentum Functions .. . . . . . . . . . . . . . . .302 D.3 Relations for the Alpha and Gamma Matrices . . . . . . .. . . . . . . . . . . . . . . . .302 Reference .. . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .303

E

Lagrangian Field Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .305 E.1 Classical Mechanics .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .305 E.1.1 Electron in External Field . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .307 E.2 Classical Field Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .308 E.3 Dirac Equation in Lagrangian Formalism .. . . . . . . . . . .. . . . . . . . . . . . . . . . .309 References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .310

Contents

xvii

F

Semiclassical Theory of Radiation .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .311 F.1 Classical Electrodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .311 F.1.1 Maxwell’s Equations in Covariant Form . . .. . . . . . . . . . . . . . . . .311 F.1.2 Coulomb Gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .315 F.2 Quantized Radiation Field . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .317 F.2.1 Transverse Radiation Field . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .317 F.2.2 Breit Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .318 F.2.3 Transverse Photon Propagator . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .321 F.2.4 Comparison with the Covariant Treatment .. . . . . . . . . . . . . . . . .322 References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .324

G

Covariant Theory of Quantum ElectroDynamics .. . . . . . .. . . . . . . . . . . . . . . . .325 G.1 Covariant Quantization: Gupta–Bleuler Formalism .. . . . . . . . . . . . . . . . .325 G.2 Gauge Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .327 G.2.1 General .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .327 G.2.2 Covariant Gauges .. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .328 G.2.3 Noncovariant Gauge .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .329 G.3 Gamma Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .332 G.3.1 z D 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .333 G.3.2 z D 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .333 References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .334

H

Feynman Diagrams and Feynman Amplitude. . . . . . . . . . . .. . . . . . . . . . . . . . . . .335 H.1 Feynman Diagrams .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .335 H.1.1 S-Matrix .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .335 H.1.2 Green’s Function.. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .336 H.1.3 Covariant Evolution Operator .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .336 H.2 Feynman Amplitude.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .337

I

Evaluation Rules for Time-Ordered Diagrams. . . . . . . . . . .. . . . . . . . . . . . . . . . .341 I.1 Single-Photon Exchange .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .342 I.2 Two-Photon Exchange . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .343 I.2.1 No Virtual Pair .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .344 I.2.2 Single Hole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .345 I.2.3 Double Holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .346 I.3 General Evaluation Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .347 References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .348

J

Some Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .349 J.1 Feynman Integrals .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .349 R d 3 k eikr12 J.2 Evaluation of the Integral .2/ 3 q 2 k 2 Ci . . . . . . . . .. . . . . . . . . . . . . . . . .351 ikr R d 3k J.3 Evaluation of the Integral .2/3 ˛1 kO ˛2 kO q 2ek212Ci . . . . . .352 References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .354

xviii

K

Contents

Unit Systems and Dimensional Analysis . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .355 K.1 Unit Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .355 K.1.1 SI System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .355 K.1.2 Relativistic or “Natural” Unit System . . . . . .. . . . . . . . . . . . . . . . .355 K.1.3 Hartree Atomic Unit System .. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .356 K.1.4 cgs Unit Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .357 K.2 Dimensional Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .357

Abbreviations .. . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .361 Index . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .363

Chapter 1

Introduction

1.1 Standard Many-Body Perturbation Theory The quantum-mechanical treatment of many-electron systems, based on the Schrödinger equation and the Coulomb interaction between the electrons, was developed shortly after the advent of quantum mechanics, particularly by John Slater in the late 1920s and early 1930s [58]. Self-consistent-field (SCF) schemes were early developed by Slater, Hartree, Fock, and others.1 Perturbative schemes for quantum-mechanical system, based on the Rayleigh–Schrödinger and Brillouin– Wigner schemes, were developed in the 1930s and 1940s, leading to the important linked-diagram expansion, introduced by Brueckner [13] and Goldstone [28] in the 1950s, primarily for nuclear applications. That scheme was in the 1960s and 1970s also applied to electronic systems [31] and extended to degenerate and quasi-degenerate energy levels [10, 35]. The next step in this development was the introduction of “all-order methods” of coupled-cluster type, where certain effects are taken to all orders of the perturbation expansion. This represents the last – and probably final – major step of the development of a nonrelativistic many-body perturbation theory (MBPT).2 The first step toward a relativistic treatment of manyelectron systems was taken in the early 1930s by Breit [11], extending works made somewhat earlier by Gaunt [25]. Physically, the Gaunt interaction represents the magnetic interaction between the electrons, which is a purely relativistic effect. Breit augmented this treatment by including the leading retardation effect, due to the fact that the Coulomb interaction is not instantaneous, which is an effect of the same order. A proper relativistic theory should be Lorentz covariant, like the Dirac singleelectron theory.3 The Dirac equation for the individual electrons together with the

1

For a review of the SCF methods, the reader is referred to the book by Froese-Fischer [24]. By MBPT, we understand here perturbative methods based upon the Rayleigh–Schrödinger perturbation scheme and the linked-diagram expansion. To that group, we also include nonperturbative schemes, such as the coupled-cluster approach (CCA), which are based upon the same formalism. 3 A physical quantity (scalar, vector, tensor) is said to be Lorentz covariant, if it transforms according to a representation of the Lorentz group. (Only a scalar is invariant under that transformation.) 2

I. Lindgren, Relativistic Many-Body Theory: A New Field-Theoretical Approach, Springer Series on Atomic, Optical, and Plasma Physics 63, c Springer Science+Business Media, LLC 2011 DOI 10.1007/978-1-4419-8309-1 1,

1

2

1 Introduction

instantaneous Coulomb and Breit interactions between the electrons represents for a many-electron system all effects up to order ˛ 2 H(artree atomic units) or ˛ 4 me c 2 .4 This procedure, however, is NOT Lorentz covariant, and the Breit interaction can only be treated to first order in perturbation theory, unless projection operators are introduced to prevent the intermediate states from falling into the “Dirac sea” of negative-energy states, as discussed early by Brown and Ravenhall [12] and later by Sucher [62]. The latter approach has been successfully employed for a long time in relativistic many-body calculations and is known as the no-virtual-pair approximation (NVPA). A fully covariant relativistic many-body theory requires a field-theoretical approach, i.e., the use of quantum-electrodynamics (QED). In principle, there is no sharp distinction between relativity and QED, but conventionally we shall refer to effects beyond the NVPA as QED effects. This includes effects of retardation, virtual pairs, and radiative effects (self-energy, vacuum polarization, vertex correction). The systematic treatment of these effects requires a covariant approach, where the QED effects are included in the wave function. The main purpose of this book is to formulate the foundations of such a procedure.

1.2 Quantum-Electrodynamics Already in the 1930s deviations were observed between the results of precision spectroscopy and the Dirac theory for simple atomic systems, primarily the hydrogen atom. Originally, this deviation was expected to be due to vacuum polarization, i.e., spontaneous creation of electron–positron pairs in the vacuum, but this effect turned out to be too small and even of the wrong sign. An alternative explanation was the electron self-energy, i.e., the emission and absorption of a virtual photon on the same electron – another effect that is not included in the Dirac theory. Early attempts to calculate this effect, however, were unsuccessful, due to singularities (infinities) in the mathematical expressions. The first experimental observation of a clear-cut deviation from the Dirac theory was the detection in 1947 by Lamb and Retherford of the so-called Lamb shift [34], namely the shift between the 2s and 2p1=2 levels in atomic hydrogen, levels that are exactly degenerate in the Dirac theory [17, 18]. In the same year, Bethe was able to explain the shift by a nonrelativistic calculation, eliminating the singularity of the self-energy by means of a renormalization process [5]. At the same time, Kusch and Foley observed that the magnetic g-factor of the free electron deviates slightly but significantly from the Dirac value 2 [32, 33]. These observations led

An equation or a theory, like the theory of relativity or Maxwell’s theory of electromagnetism, is said to be Lorentz covariant, if it can be expressed entirely in terms of covariant quantities (see, for instance, the books of Bjorken and Drell [7, 8]). 4 ˛ is the fine-structure constant 1=137 and me c 2 is the electron rest energy (see Appendix K).

1.3 Bethe–Salpeter Equation

3

to the development of the modern form of the quantum-electrodynamic theory by Feynman, Schwinger, Dyson, Tomanaga, and others by which the deviations from the Dirac theory could be explained with good accuracy [20, 22, 23, 56, 64].5 The original theory of QED was applied to free electrons. During the last four decades, several methods have been developed for numerical calculation of QED effects in bound electronic states. The scattering-matrix or S-matrix formulation, originally developed for dealing with the scattering of free particles, was made applicable also to bound states by Sucher [60], and the numerical procedure was refined in the 1970s particularly by Mohr [38]. During the last two decades, the method has been extensively used in studies of highly charged ions to test the QED theory under extreme conditions, works that have been pioneered by Mohr and Soff (for a review, see [39]). The Green’s function is one of the most important tools in mathematical physics with applications in essentially all branches of physics.6 In 1990s, the method was adopted to bound-state QED problems by Shabaev et al. [57]. This procedure is referred to as the Two-times Green’s function and has recently been extensively applied to highly charged ions by the St Petersburg group. During the first decade of this century, another procedure for numerical QED calculations was developed by the Gothenburg atomic theory group, termed the Covariant-evolution-operator (CEO) method [36], which has been applied to the fine structure and other energy-level separations of helium-like ions.

1.3 Bethe–Salpeter Equation The first completely covariant treatment of a bound-state problem was presented in 1951 by Salpeter and Bethe [6, 52] and by Gell-Mann and Low [26]. The Bethe–Salpeter (BS) equation contains in principle the complete relativistic and interelectronic interaction, i.e., all kinds of electron correlation and QED effects. The BS equation is associated with several fundamental problems, which were discussed in the early days, particularly by Dyson [21], Goldstein [27], Wick [65], and Cutkosky [16]. Dyson found that the question of relativistic quantum mechanics is “full of obscurities and unsolved problems” and that “the physical meaning of the four-dimensional wave function is quite unclear.” It seems that some of these problems still remain. The BS equation is based upon field theory, and there is no direct connection to the Hamiltonian approach of relativistic quantum mechanics. The solution of the field-theoretical BS equation leads to a four-dimensional wave function with individual times for the two particles. This is not in accordance with the standard 5

For the history of the development of the QED theory, the reader is referred to the authoritative review by Schweber [55]. 6 For a comprehensive account of the applications, particularly in condensed-matter physics, the reader is referred to the book by Mahan [37].

4

1 Introduction

quantum-mechanical picture, which has a single time variable also for many-particle systems. The additional time variable leads sometimes to “abnormal solutions” with no counterparts in nonrelativistic quantum mechanics, as discussed particularly by Nakanishi [40] and Namyslowski [41]. Much efforts have been devoted to simplifying the BS equation by reducing it to a three-dimensional equation, in analogy with the standard quantum-mechanical equations (for reviews, see [9, 15]). Salpeter [51] derived early an “instantaneous” approximation, neglecting retardation, which led to a relativistically exact threedimensional equation, similar to – but not exactly equal to – the Breit equation. More sophisticated is the so-called quasi-potential approximation, introduced by Todorov [63], frequently used in scattering problems. Here, a three-dimensional Schrödinger-type equation is derived with an energy-dependent potential, deduced from scattering theory. Sazdjian [53, 54] was able to separate the BS equation into a three-dimensional equation of Schrödinger type and one equation for the relative time of the two particles, serving as a perturbation – an approach that is claimed to be exactly equivalent to the original BS equation. This approach establishes a definitive link between the Hamiltonian relativistic quantum mechanics and field theory. Connell [15] further developed the quasi-potential approximation of Todorov by introducing series of corrections, a procedure that also is claimed to be formally equivalent to the original BS equation. Caswell and Lepage [14] applied the quasi-potential method to evaluate the hyperfine structure of muonium and positronium to the order ˛ 6 me c 2 by combining analytical and perturbative approaches. Grotch and Yennie [9, 30] have applied the method to evaluate higher-order nuclear corrections to the energy levels of the hydrogen atom, and Adkins and Fell [1, 2] have applied it to positronium. The procedure we shall develop in the following is to combine the covariantevolution-operator method with electron correlation, which will constitute a step toward a fully covariant treatment of many-electron systems. This will form another approximation of the full Bethe–Salpeter equation that seems feasible for electronic systems. A vast literature on the Bethe–Salpeter equation, its fundamental problems and its applications, has been gathered over the years since the original equation appeared. Most applications are performed in the strong-coupling case (QCD), where the fundamental problems of the equation are more pronounced. The interested reader is here referred to some reviews of the field, where numerous references to original works can be found [29, 40–42, 54].

1.4 Helium Atom: Analytical Approach An approach to solve the BS equation, known as the external-potential approach, was first developed by Sucher [59, 61] to evaluate the lowest-order QED contributions to the ground-state energy of the helium atom, and equivalent results were at the same time also derived by Araki [3]. The electrons are here assumed to move

1.5 Field-Theoretical Approach to Many-Body Perturbation Theory

5

in the field of the (infinitely heavy) atomic nucleus. The relative time of the two electrons is eliminated by integrating over the corresponding energy of the Fourier transform, which leads to a Schrödinger-like equation, as in the quasi-potential method. The solution of this equation is expanded in terms of a Brillouin–Wigner perturbation series. This work has been further developed and applied by Douglas and Kroll [19] and by Zhang and Drake [69, 70] by considering higher-order terms in the ˛ and Z˛ expansions. This approach, which is reviewed in Chap. 11, can be used for light systems, such as light helium-like ions, where the power expansions are sufficiently convergent. The QED effects are here evaluated by means of highly correlated wave functions of Hylleraas type, which implies that QED and electroncorrelation effects are highly mixed. A related technique, referred to as the effective Hamiltonian approach, has been developed and applied to helium-like systems by Pachucki and Sapirstein [43–45]. A problem that has been controversial for quite some time is the fine structure of the lowest P state of the neutral helium atom. The very accurate analytical results of Drake et al. and by Pachucki et al. give results close to the experimental results obtained by Gabrielse and others [68], but there have for quite some time been significant deviations – well outside the estimated limits of error. Very recently, Pachucki and Yerokhin have by means of improved calculations shown that the controversy has been resolved [46, 47, 66, 67].

1.5 Field-Theoretical Approach to Many-Body Perturbation Theory The methods previously mentioned for numerical QED calculations can for computational reasons be applied only to one- and two-photon exchange, which implies that the electron correlation is treated at most to second order. This might be sufficiently accurate for highly charged systems, where the QED effects dominate over the electron correlation, but is usually quite insufficient for lighter systems, where the situation is reversed. To remedy the situation to some extent, higher-order manybody contributions can be added to the two-photon energy, a technique applied by the Gothenburg and St Petersburg groups [4, 48]. In the numerical procedures for standard (relativistic) MBPT, the electron correlation can be evaluated effectively to essentially all orders by technique of coupledcluster type. QED effects can here be included only as first-order energy corrections, a technique applied particularly by the Notre-Dame group [49]. To treat electron correlation, relativity and QED in a unified manner would require a field-theoretical approach. The above-mentioned methods for QED calculations are all based upon fieldtheory. Of these methods, the covariant-evolution method has the advantage that it has a structure that is quite akin to that of standard MBPT, which has the consequence that it can serve as a basis for a unified field-theoretical many-body approach. The QED effects can here be included in the wave function, which will

6

1 Introduction

make it possible to treat the QED and correlation effects in a more unified way. To solve this problem completely is a formidable task, but it will be a main theme of this book to describe how some steps can be taken in this direction, along the line that is presently being pursued by the Gothenburg atomic theory group. The covariant evolution operator, which describes the time evolution of the relativistic state vector, is the key tool in this treatment. This operator is closely related to the field-theoretical Green’s function. It should be mentioned that a related idea was proposed by Leonard Rosenberg already 20 years ago [50], namely of including Coulomb interactions in the QED Hamiltonian, and this is essentially the procedure we are pursuing in this book. The covariant evolution operator is singular, as is the standard evolution operator of nonrelativistic quantum mechanics, but the singularities can be eliminated in a similar way as the corresponding singularities of the Green’s function. The regular part of the covariant evolution operator is referred to as the Green’s operator, which can be regarded as an extension of the Green’s-function concept and shown to serve as a link between field theory and standard many-body perturbation theory. The perturbation used in this procedure represents the interaction between the electromagnetic field and the individual electrons. This implies that the equations operate in an extended photonic Fock space with variable number of photons. The strategy in dealing with the combined QED and correlation problem is first to construct a field-theoretical “QED potential” with a single retarded photon, containing all first-order QED effects (retardation, virtual pairs, radiative effects), which – after proper regularization and renormalization – can be included in a perturbative expansion of MBPT or coupled-cluster type. In this way, the QED effects can – for the first time – be built into the wave function and treated together with the electron correlation in a coherent manner. For practical reasons, only a single retarded photon (together with arbitrary number of Coulomb interactions) can be included in this procedure at present time, but due to the fact that these effects are included in the wave function, this corresponds to higher-order effects in the energy. When extended to interactions of multiphoton type, this leads for two-particle systems to the Bethe–Salpeter equation, and in the multireference case to an extension of this equation, referred to as the Bethe–Salpeter–Bloch equation. In combining QED with electron correlation, it is necessary to work in the Coulomb gauge, in order to take advantage of the development in standard MBPT. Although this gauge is noncovariant in contrast to, for instance, the simpler Feynman gauge, it can be argued that the deviation from a fully covariant treatment will have negligible effect in practical applications when handled properly. This makes it possible to mix a larger number of Coulomb interactions with the retarded-photon interactions, which is expected to lead to the same ultimate result as a fully covariant approach but with faster convergence rate due to the dominating role of the Coulomb interaction. The procedure can also be extended to systems with more than two electrons, and due to the complete compatibility between the standard and the extended procedures, the QED effects need only be included where they are expected to be most significant.

References

7

In principle, also the procedure outlined here leads to individual times for the particles involved, consistent with the full Bethe–Salpeter equation but not with the standard quantum-mechanical picture. We shall mainly work in the equal-time approximation here, and we shall not analyze effects beyond this approximation in any detail. It is expected that – if existing – any such effect would be extremely small for electronic systems.

References 1. Adkins, G.S., Fell, R.N.: Bound-state formalism for positronium. Phys. Rev. A 60, 4461–75 (1999) 2. Adkins, G.S., Fell, R.N., Mitrikov, P.M.: Calculation of the positronium hyperfine interval using the Bethe-Salpeter formalism. Phys. Rev. A 65, 042,103 (2002) 3. Araki, H.: Quantum-Electrodynamical Corrections to Energy-levels of Helium. Prog. Theor. Phys. (Japan) 17, 619–42 (1957) 4. Artemyev, A.N., Shabaev, V.M., Yerokhin, V.A., Plunien, G., Soff, G.: QED calculations of the n D 1 and n D 2 energy levels in He-like ions. Phys. Rev. A 71, 062,104 (2005) 5. Bethe, H.A.: The Electromagnetic Shift of Energy Levels. Phys. Rev. 72, 339–41 (1947) 6. Bethe, H.A., Salpeter, E.E.: An Introduction to Relativistic Quantum Field Theory. Quantum Mechanics of Two-Electron Atoms. Springer, Berlin (1957) 7. Bjorken, J.D., Drell, S.D.: Relativistic Quantum Fields. Mc-Graw-Hill Pbl. Co, N.Y. (196) 8. Bjorken, J.D., Drell, S.D.: Relativistic Quantum Mechanics. Mc-Graw-Hill Pbl. Co, N.Y. (1964) 9. Boldwin, G.T., Yennie, D.R., Gregorio, M.A.: Recoil effects in the hyperfine structure of QED bound states. Rev. Mod. Phys. 57, 723–82 (1985) 10. Brandow, B.H.: Linked-Cluster Expansions for the Nuclear Many-Body Problem. Rev. Mod. Phys. 39, 771–828 (1967) 11. Breit, G.: Dirac’s equation and the spin-spin interaction of two electrons. Phys. Rev. 39, 616–24 (1932) 12. Brown, G.E., Ravenhall, D.G.: On the Interaction of Two electrons. Proc. R. Soc. London, Ser. A 208, 552–59 (1951) 13. Brueckner, K.A.: Many-Body Problems for Strongly Interacting Particles. II. Linked Cluster Expansion. Phys. Rev. 100, 36–45 (1955) 14. Caswell, W.E., Lepage, G.P.: Reduction of the Bethe-Salpeter equation to an equivalent Schrödinger equation, with applications. Phys. Rev. A 18, 810–19 (1978) 15. Connell, J.H.: QED test of a Bethe-Salpeter solution method. Phys. Rev. D 43, 1393–1402 (1991) 16. Cutkosky, R.E.: Solutions of the Bethe-Salpeter equation. Phys. Rev. 96, 1135–41 (1954) 17. Dirac, P.A.M.: . Roy. Soc. (London) 117, 610 (1928) 18. Dirac, P.A.M.: The Principles of Quantum Mechanics. Oxford Univ. Press, Oxford (1930, 1933, 1947, 1958) 19. Douglas, M.H., Kroll, N.M.: Quantum Electrodynamical Corrections to the Fine Structure of Helium. Ann. Phys. (N.Y.) 82, 89–155 (1974) 20. Dyson, F.J.: The radiation Theories of Tomonaga, Schwinger, and Feynman. Phys. Rev. 75, 486–502 (1949) 21. Dyson, F.J.: The Wave Function of a Relativistic System. Phys. Rev. 91, 1543–50 (1953) 22. Feynman, R.P.: Space-Time Approach to Quantum Electrodynamics. Phys. Rev. 76, 769–88 (1949) 23. Feynman, R.P.: The Theory of Positrons. Phys. Rev. 76, 749–59 (1949) 24. Froese-Fischer, C.: The Hartree-Fock method for atoms. John Wiley and Sons, New York, London, Sidney, Toronto (1977)

8

1 Introduction

25. Gaunt, J.A.: The Triplets of Helium. Proc. R. Soc. London, Ser. A 122, 513–32 (1929) 26. Gell-Mann, M., Low, F.: Bound States in Quantum Field Theory. Phys. Rev. 84, 350–54 (1951) 27. Goldstein, J.S.: Properties of the Salpeter-Bethe Two-Nucleon Equation. Phys. Rev. 91, 1516–24 (1953) 28. Goldstone, J.: Derivation of the Brueckner many-body theory. Proc. R. Soc. London, Ser. A 239, 267–279 (1957) 29. Grotch, H., Owen, D.A.: Bound states in Quantum Electrodynamics: Theory and Applications. Fundamentals of Physics 32, 1419–57 (2002) 30. Grotch, H., Yennie, D.R.: Effective Potential Model for Calculating Nuclear Corrections to the Eenergy Levels of Hydrogen. Rev. Mod. Phys. 41, 350–74 (1969) 31. Kelly, H.P.: Application of many-body diagram techniques in atomic physics. Adv. Chem. Phys. 14, 129–190 (1969) 32. Kusch, P., Foley, H.M.: Precision Measurement of the Ratio of the Atomic ’g Values’ in the 2 P3=2 and 2 P1=2 States of Gallium. Phys. Rev. 72, 1256–57 (1947) 33. Kusch, P., Foley, H.M.: On the Intrinsic Moment of the Electron. Phys. Rev. 73, 412 (1948) 34. Lamb, W.W., Retherford, R.C.: Fine structure of the hydrogen atom by microwave method. Phys. Rev. 72, 241–43 (1947) 35. Lindgren, I.: The Rayleigh-Schrödinger perturbation and the linked-diagram theorem for a multi-configurational model space. J. Phys. B 7, 2441–70 (1974) ˚ en, B.: The covariant-evolution-operator method in bound36. Lindgren, I., Salomonson, S., As´ state QED. Physics Reports 389, 161–261 (2004) 37. Mahan, G.D.: Many-particle Physics, second edition. Springer Verlag, Heidelberg (1990) 38. Mohr, P.J.: Numerical Evaluation of the 1s1=2 -State Radiative Level Shift. Ann. Phys. (N.Y.) 88, 52–87 (1974) 39. Mohr, P.J., Plunien, G., Soff, G.: QED corrections in heavy atoms. Physics Reports 293, 227–372 (1998) 40. Nakanishi, N.: Normalization condition and normal and abnormal solutions of Bethe-Salpeter equation. Phys. Rev. 138, B1182 (1965) 41. Namyslowski, J.M.: The Relativistic Bound State Wave Function. in Light-Front Quantization and Non-Perturbative QCD, J.P. Vary and F. Wolz, eds. (International Institute of Theoretical and Applied Physics, Ames) (1997) 42. Onida, G., Reining, L., Rubio, A.: Electronic excitations: density-functional versus many-body Green’s-function approaches. Rev. Mod. Phys. 74, 601–59 (2002) 43. Pachucki, K.: Quantum electrodynamics effects on helium fine structure. J. Phys. B 32, 137–52 (1999) 44. Pachucki, K.: Improved Theory of Helium Fine Structure. Phys. Rev. Lett. 97, 013,002 (2006) 45. Pachucki, K., Sapirstein, J.: Contributions to helium fine structure of order m˛ 7 . J. Phys. B 33, 5297–5305 (2000) 46. Pachucky, K., Yerokhin, V.A.: Reexamination of the helium fine structure (vol 79, 062516, 2009). Phys. Rev. Lett. 80, 19,902 (2009) 47. Pachucky, K., Yerokhin, V.A.: Reexamination of the helium fine structure (vol 79, 062516, 2009). Phys. Rev. A 81, 39,903 (2010) 48. Persson, H., Salomonson, S., Sunnergren, P., Lindgren, I.: Two-electron Lamb-Shift Calculations on Heliumlike Ions. Phys. Rev. Lett. 76, 204–07 (1996) 49. Plante, D.R., Johnson, W.R., Sapirstein, J.: Relativistic all-order many-body calculations of the n D 1 and n D 2 states of heliumlike ions. Phys. Rev. A 49, 3519–30 (1994) 50. Rosenberg, L.: Virtual-pair effects in atomic structure theory. Phys. Rev. A 39, 4377–86 (1989) 51. Salpeter, E.E.: Mass Correction to the Fine Structure of Hydrogen-like Atoms. Phys. Rev. 87, 328–43 (1952) 52. Salpeter, E.E., Bethe, H.A.: A Relativistic Equation for Bound-State Problems. Phys. Rev. 84, 1232–42 (1951) 53. Sazdjian, H.: Relativistiv wave equations for the dynamics of two interacting particles. Phys. Rev. D 33, 3401–24 (1987)

References

9

54. Sazdjian, H.: The connection of two-particle relativistic quantum mechanics with the BetheSalpeter equation. J. Math. Phys. 28, 2618–38 (1987) 55. Schweber, S.S.: The men who madt it: Dyson, Feynman, Schwinger and Tomonaga. Princeton University Press, Princeton (1994) 56. Schwinger, J.: Quantum electrodynamics I. A covariant formulation. Phys. Rev. 74, 1439 (1948) 57. Shabaev, V.M.: Two-times Green’s function method in quantum electrodynamics of high-Z few-electron atoms. Physics Reports 356, 119–228 (2002) 58. Slater, J.: Quantum Theory of Atomic Spectra. McGraw-Hill, N.Y. (1960) 59. Sucher, J.: Energy Levels of the Two-Electron Atom to Order ˛ 3 Ry; Ionization Energy of Helium. Phys. Rev. 109, 1010–11 (1957) 60. Sucher, J.: S-Matrix Formalism for Level-Shift Calculations. Phys. Rev. 107, 1448–54 (1957) 61. Sucher, J.: Ph.D. thesis, Columbia University (1958). Univ. Microfilm Internat., Ann Arbor, Michigan 62. Sucher, J.: Foundations of the Relativistic Theory of Many Electron Atoms. Phys. Rev. A 22, 348–62 (1980) 63. Todorov, I.T.: Quasipotential Equation Corresponding to the relativistic Eikonal Approximation. Phys. Rev. D 3, 2351–56 (1971) 64. Tomanaga, S.: On Infinite Field Reactions in Quantum Field Theory. Phys. Rev. 74, 224–25 (1948) 65. Wick, G.C.: Properties of Bethe-Salpeter Wave Functions. Phys. Rev. 96, 1124–34 (1954) 66. Yerokhin, K.P.V.A.: Reexamination of the helium fine structure. Phys. Rev. Lett. 79, 62,616 (2009) 67. Yerokhin, K.P.V.A.: Fine Structure of Heliumlike Ions and Determination of the Fine Structure Constant. Phys. Rev. Lett. 104, 70,403 (2010) 68. Zelevinsky, T., Farkas, D., Gabrielse, G.: Precision Measurement of the Three 23 PJ Helium Fine Structure Intervals. Phys. Rev. Lett. 95, 203,001 (2005) 69. Zhang, T.: Corrections to O.˛ 7 .ln ˛/mc 2 / fine-structure splittings and O.˛ 6 .ln ˛/mc 2 / energy levels in helium. Phys. Rev. A 54, 1252–1312 (1996) 70. Zhang, T., Drake, G.W.F.: Corrections to O.˛ 7 mc 2 / fine-structure splitting in helium. Phys. Rev. A 54, 4882–4922 (1996)

Part I

Basics: Standard Many-Body Perturbation Theory

Chapter 2

Time-Independent Formalism

In this first part of the book, we shall review some basics of quantum mechanics and the many-body theory for bound electronic systems that will form the foundations for the following treatment. This material can also be found in several standard text books. The time-independent formalism is summarized in the present chapter1 and the time-dependent formalism in the following one.

2.1 First Quantization First quantization is the term for the elementary treatment of quantized systems, where the particles of the system are treated quantum-mechanically, for instance, in terms of Schrödinger wave functions, while the surrounding fields are treated classically.

2.1.1 De Broglie’s Relations As an introduction to the quantum mechanics, we shall derive the Schrödinger equation from the classical relations of de Broglie. According to Planck–Einstein’s quantum theory, the electromagnetic radiation is associated with particle-like photons with the energy (E) and momentum (p) given by the relations

E D h D !„ ; p D h= D „k

(2.1)

1

This chapter is essentially a short summary of the second part of the book Atomic Many-Body Theory by Lindgren and Morrison, and the reader who is not well familiar with the subject is recommended to consult that book.


13

14

2 Time-Independent Formalism

where „ D h=2, h being Planck’s constant (see further Appendix K), the cyclic frequency of the radiation (cycles/s) and ! D 2 the angular frequency (radians/s). D c= (c being the velocity of light in vacuum) is the wavelength of the radiation and k D 2= the wave number. De Broglie assumed that the relations (2.1) for photons would hold also for material particles, like electrons. Nonrelativistically, we have for a free electron in one dimension p2 „2 k 2 ED or „! D ; (2.2) 2me 2me where me is the mass of the electron. De Broglie assumed that a particle could be represented by a wave packet Z .t; x/ D dk a.k/ ei.kx!t / : (2.3) The relation (2.2) then leads to the one-dimensional wave equation for a free electron „2 @2 .t; x/ @.t; x/ D i„ ; (2.4) @t 2me @x 2 which is the Schrödinger equation for a free particle. This can be obtained from the first of the relations (2.2) by means of the substitutions E ! i„

@ @t

p ! i„

@ : @x

(2.5)

2.1.2 The Schrödinger Equation We can generalize the treatment above to an electron in three dimensions in an external field, vext .x/, for which the energy Hamiltonian is EDH D

p2 C vext .x/: 2me

(2.6)

Generalizing the substitutions above to2 p ! pO D i„r

and

x ! xO D x;

(2.7)

where r is the vector gradient operator (see Appendix A.1) leads to the Hamilton operator pO 2 „2 2 HO D C vext .x/ D r C vext .x/ (2.8) 2me 2me 2

Initially, we shall use the ‘hat’ symbol to indicate an operator, but later we shall use this symbol only when the operator character needs to be emphasized.

2.1 First Quantization

15

and to the Schrödinger equation for a single electron @ „2 2 O i„ .t; x/ D H .t; x/ D r C vext .x/ .t; x/: @t 2me

(2.9)

For an N -electron system, the Schrödinger equation becomes correspondingly3 i„

@ .tI x 1 ; x 2 ; x N / D HO .tI x 1 ; x 2 ; x N /; @t

(2.10)

where we assume the Hamiltonian to be of the form HO D HO 1 C HO 2 [see Appendix (B.19)]4 HO 1 D

N N X X „2 2 hO 1 .n/; rn C vext .x n / DW 2me

nD1

HO 2 D

N X

nD1

N X

e2 DW hO 2 .m; n/: 40 rmn m t2 : t1 < t2 (2.31)

Here, O ˙ represents the positive-/negative-energy part of the spectrum, respectively, and p and h denote particle (positive-energy) and hole (negative-energy) states, respectively. The results can be summarized as O .x1 / O .x2 / D O .x2 / O .x1 / D j .x 1 / j .x 2 / ei"j .t1 t2 /=„ ;

(2.32)

if t1 > t2 for particles and t1 < t2 for holes with all other contractions vanishing.

2.2.3 Wick’s Theorem The handling of operators in second quantization is greatly simplified by Wick’s theorem [80] (for an introduction, see, for instance, Fetter and Walecka [19, Sect. 8] or Lindgren and Morrison [40, Chap. 11]), which states that a product of creation and annihilation operators AO can be written as the normal product plus all single, double . . . contractions with the uncontracted operators in normal form, or symbolically O C fAg: O AO D fAg

(2.33)

A particularly useful form of Wick’s theorem is the following. If AO and BO are operators in normal form, then the product is equal to the normal product plus all O or formally normal-ordered contractions between AO and B, O C fAO Bg: O AO BO D fAO Bg

(2.34)

With this formulation, there are no further contractions within the operators to be multiplied. This forms the basic rule for the graphical representation of the operators and operator relations to be discussed below.

20


2.3 Time-Independent Many-Body Perturbation Theory 2.3.1 Bloch Equation Here, we shall summarize the most important concepts of standard timeindependent many-body perturbation theory (MBPT) as a background for the further treatment. (For more details, the reader is referred to designated books, such as Lindgren–Morrison, Atomic Many-Body Theory [40].) We are considering a number of stationary electronic states,j ˛ i .˛ D 1 d /, termed target states, that satisfy the Schrödinger equation Hj ˛ i D E ˛j ˛ i

.˛ D 1 d /:

(2.35) ˛ ˛

For each target state, there exists an “approximate” or model state, j0 .˛ D 1 d /, which is more easily accessible and which forms the starting point for the perturbative treatment. We assume that the model states are linearly independent and that they span a model space. The projection operator for the model space is denoted P and that for the complementary or orthogonal space by Q, which together form the identity operator P C Q D I:

(2.36)

A wave operator is introduced – also known as the Møller operator [53] – which transforms the model states back to the exact states, j ˛ i D ˝j0˛ i

.˛ D 1 d /

(2.37)

and this operator is the same for all states under consideration. We define an effective Hamiltonian with the property that operating on a model function it generates the corresponding exact energy Heffj0˛ i D E ˛j0˛ i

.˛ D 1 d /;

(2.38)

with the eigenvectors representing the model states. Operating on this equation with ˝ from the left, using the definition (2.37), yields ˝Heffj0˛ i D E ˛j ˛ i ;

(2.39)

which we compare with the Schrödinger equation (2.35) H ˝j0˛ i D E ˛j ˛ i :

(2.40)

Since this relation holds for each state of the model space, we have the important operator relation ˝Heff P D H ˝P; which as known as the generalized Bloch equation.

(2.41)

2.3 Time-Independent Many-Body Perturbation Theory

21

The form above of the Bloch equation is valid independently on the choice of normalization. In the following, we shall mainly work with the intermediate normalization (IN), which implies h0˛j ˛ i D 1; j0˛ i D P j ˛ i

.˛ D 1 d /:

(2.42a) (2.42b)

Then we have after projecting the Schrödinger equation onto the model space PH ˝j0˛ i D E ˛j0˛ i

(2.43)

and we find that the effective Hamiltonian (2.38) becomes in IN Heff D PH ˝P:

(2.44)

Normally, the multidimensional or multireference model space is applied in connection with valence universality, implying that the same operators are used for different stages of ionization (see further Sect. 2.5).

2.3.2 Partitioning of the Hamiltonian For electrons moving in an external (nuclear) potential, vext , the single-electron (Schrödinger) Hamiltonian (2.8) is hS D

„2 2 r C vext : 2me

(2.45)

The corresponding Schrödinger equation hS i .x/ D "i i .x/

(2.46)

generates a complete spectrum of functions, which can form the basis for numerical calculations. This is known to as the Furry picture. These single-electron functions are normally referred to as (single-electron) orbitals – or spin-orbitals, if a spin eigenfunction is adhered. Degenerate orbitals (with the same eigenvalue) form an electron shell. The Hamiltonian for a many-electron system (2.11) is N N X X e2 „2 2 H D r C vext C ; 2me 40 rnm n n n<m

(2.47)

22


where the last term represents the interelectronic interaction. For the perturbation treatment, we separate the many-electron Hamiltonian into H D H0 C V;

(2.48)

where H0 a model Hamiltonian that is a sum of single-electron Hamiltonians N N X X „2 2 r C vext C u DW h0 .n/ H0 D 2me n n n

(2.49)

and V is a perturbation V D

N X

un C

n

N X

e2 : 40 rnm n<m

(2.50)

The potential u is optional and used primarily to improve the convergence properties of the perturbation expansion. The antisymmetrized N -electron eigenfunctions of H0 can be expressed as determinantal products of single-electron orbitals (see Appendix B) H0 ˚A .x 1 ; x 2 x N / D E0A ˚A .x 1 ; x 2 x N /; p ˚A .x 1 ; x 2 x N / D 1= N Š Af 1 .x 1 / 2 .x 2 / N .x N /g;

(2.51)

where A is an antisymmetrizing operator. The determinants are referred to as Slater determinants and constitute our basis functions. The eigenvalues are given by E0 D

N X

"n

(2.52)

nD1

summed over the spin-orbitals of the determinant. Degenerate determinants form a configuration. The model space is supposed to be formed by one or several configurations that can have different energies. We distinguish between three kinds of orbitals Core orbitals, present in all determinants of the model space Valence orbitals, present in some determinants of the model space Virtual orbitals, not present in any determinants of the model space

The model space is said to be complete, if it contains all configurations that can be formed by distributing the valence electrons among the valence orbitals in all possible ways. In the following, we shall normally assume this to be the case. With the partitioning (2.48), the Bloch equation above can be expressed .˝Heff H0 ˝/ P D V ˝P:

(2.53)

2.3 Time-Independent Many-Body Perturbation Theory

23

With H0 of the form (2.49), it commutes with the projection operator P . Then, we find that Heff D PH0 P C P V ˝P

(2.54)

and we shall refer to the second term as the effective interaction Veff D P V ˝P:

(2.55)

The partitioning leads to the commonly used form of the generalized Bloch

equation [36, 37, 40] Œ˝; H0 P D Q .V ˝ ˝Veff / P;

(2.56)

which is frequently used as the basis for many-body perturbation theory (MBPT). The last term appears only for open-shell systems with unfilled valence shell(s) and is graphically represented by so-called folded or backwards diagrams, first introduced by Brandow in nuclear physics [7], (see further below). If the model space is completely degenerate with a single energy E0 , the general Bloch equation reduces to its original form, derived in the late 1950s by Claude Bloch [4, 5], .E0 H0 / ˝P D V ˝P ˝Veff :

(2.57)

This equation can be used to generate the standard Rayleigh–Schrödinger perturbation expansion, found in many text books. The generalized Bloch equation (2.56) is valid for a general model space, which can contain different zeroth-order energy levels. Using such an extended model space represents usually a convenient way of treating very closely spaced or quasidegenerate unperturbed energy levels, a phenomenon that otherwise can lead to serious convergence problems. This can be illustrated by the relativistic calculation of the fine structure of helium-like ions, where a one-dimensional model space leads to convergence problems for light elements, a problem that can normally be remedied in a straightforward way by means of the extended model space [50, 62]. But the extended model space can also lead to problems, due to so-called intruder states, as will be further discussed below. With an extended model space, we can separate the projection operator into the corresponding energy components5 P D

X

PE I

H0 PE D EPE :

(2.58)

E

In the case of an extended model space, we shall normally use the symbol E for the different energies of the model space.

5

24


Operating with the general Bloch equation (2.56) on a particular component, then yields .E H0 / ˝PE D Q .V ˝ ˝Veff / PE : (2.59) Expanding the wave operator order-by-order ˝ D 1 C ˝ .1/ C ˝ .2/ C

(2.60)

leads to the recursive formula

or

.E H0 / ˝ .n/ PE D Q V ˝ .n1/ .˝Veff /.n/ PE

(2.61)

˝ .n/ PE D Q .E/ V ˝ .n1/ .˝Veff /.n/ PE ;

(2.62)

.k/ Veff D P V ˝ .k1/ P:

(2.63)

where Here,

.E/ D

1 E H0

(2.64)

and

Q .E/ D Q .E/

(2.65)

are known as the resolvent and the reduced resolvent, respectively [45]. The recursive formula (2.62) can generate a generalized form of the Rayleigh– Schrödinger perturbation expansion (see [40, Chap. 9]), valid also for a quasidegenerate model space. We see from the form of the resolvent that in each new order of the perturbation expansion there is a denominator equal to the energy difference between the initial and final states. This leads to the Goldstone rules in the evaluation of the time-ordered diagrams to be considered in the following section. Even if the perturbation is energy independent, we see that the wave operator and effective interaction will still generally be energy dependent, due to the energy dependence of the resolvent. In first order, we have ˝ .1/ PE D Q .E/VPE

(2.66)

.1/ ˝ .2/ .E/PE D Q .E/ V ˝ .1/ .E/ ˝ .1/ .E 0 /PE 0 Veff PE ;

(2.67)

and in second order

.1/

where Veff D P VP . Note that the wave operator in the last term operates on the projection operator PE 0 and therefore depends on the corresponding energy E 0 . We now have ı˝ .1/ .E/ ı Q .E/

Q .E 0 / Q .E/ D V D V D Q .E/ Q .E 0 /V ıE ıE E0 E (2.68) D Q .E/˝ .1/ .E 0 /

2.4 Graphical Representation

25

and we note that the last folded term in (2.67) has a double denominator. We can express the second-order Bloch equation as ˝ .2/ .E/PE D Q .E/V ˝ .1/ .E/PE C

ı˝ .1/ .E/ .1/ Veff .E/PE : ıE

(2.69)

In the limit of complete degeneracy space, the difference ratio, of course, goes over into the partial derivative. We shall show in later chapters that the second-order expression above holds also when the perturbation is energy dependent (6.77).

2.4 Graphical Representation In this section, we shall briefly describe a way of representing the perturbation expansion graphically. (For further details, the reader is referred to the book by Lindgren and Morrison [40].)

2.4.1 Goldstone Diagrams The Rayleigh–Schrödinger perturbation expansion can be conveniently represented in terms of diagrams by means of second quantization (see above and Appendix B). The perturbation (2.50) becomes in second quantization 1 VO D ci cj hi jf jj i C ci cj cl ck hij jgjkli ; 2

(2.70)

where f is the negative potential f D u and g is the Coulomb interaction between the electrons. When some of the states above are hole states, the expression (2.70) is not in normal order. By normal ordering the expression, zero-, one-, and two-body operators will appear [40, Eq. 11.39] V D V0 C V1 C V2 ;

(2.71)

where V0 D

hole X

1X Œhij jgjkli hj i jgjkli ; 2 hole

hi jf ji i C

i

ij

V1 D

fci cj g

V2 D

1 fc c cl ck g hij jgjkli : 2 i j

hi jveffjj i ; (2.72)

26


In the one- and two-body parts, the summation is performed over all orbitals. Here, hi jveffjj i D hi jf jj i C

hole X

Œhi kjgjj ki hki jgjj ki

(2.73)

k

is known as the effective potential interaction and can be represented graphically as shown in Fig. 2.3. The summation term represents the Hartree–Fock potential hi jvHFjj i D

hole X

Œhi kjgjj ki hki jgjj ki ;

(2.74)

k

where the first term is a “direct” integral and the second term an “exchange” integral. In the Hartree–Fock model, we have u D vHF , and the effective potential vanishes [40]. We can now represent the perturbation (2.72) by the normal-ordered diagrams in Fig. 2.1. The zero- and one-body parts are shown in more detail in Figs. 2.2 and 2.3. In our diagrams, the dotted line with the cross represents the potential interaction, f D u, and the dotted line between the electrons the Coulomb interaction, g D e 2 =40 r12 . We use here a simplified version of Goldstone diagrams. Each free vertical line at the top (bottom) represents an electron creation (absorption) operator but normally we do not distinguish between the different kinds of orbitals (core, valence, and virtual) as done traditionally. There is a summation of internal lines over all orbitals of the same category. We use here heavy lines to indicate

j

V D V0 C

-u

i

i

j

k

l

+

Fig. 2.1 Graphical representation the effective-potential interaction (2.72). The heavy lines represent the orbitals in the Furry picture. The dotted line with the cross represents the potential u and the dotted, horizontal lines the Coulomb interaction. The zero-body and one-body parts of the interaction are depicted in Figs. 2.2 and 2.3, respectively

i

V0 D 6

u

k

+

6

k

l ?

+

l

Fig. 2.2 Graphical representation of the zero-body part of the effective-potential interaction (2.72). The orbitals are summed over all core/hole states


i

i

=

j

27

u

i

?+

+

j

i

j

j

Fig. 2.3 Graphical representation of the effective-potential interaction (2.73). For the closed orbital lines (with no free end), there is a summation over the core/hole states. The last two diagrams represent the “Hartree–Fock” potential, and the entire effective-potential interaction vanishes when HF orbitals are used

that the orbitals are generated in an external (nuclear) potential, i.e., the bound-state representation or Furry picture. By means of Wick’s theorem, we can now normal order the right-hand side (r.h.s.) of the perturbation expansion of the Bloch equation (2.62), and Each resulting normal-ordered term will be represented by a diagram.

The first-order wave operator (2.66) ˝ .1/ PE D Q .E/ VPE D Q .E/ .V1 C V2 /PE

(2.75)

becomes in second quantization (2.72) n

o hi jv jj i o 1n hij jgjkli eff C ˝ PE D Q c c cl ck PE : " j "i 2 i j "k C "l "i "j (2.76) This can be represented in the same way as the open part (V1 C V2 ) of the perturbation (2.70) (Fig. 2.1), if we include the extra energy denominator according to the Goldstone rules, summarized below. In second order, we have from (2.67), using Wick’s theorem (2.34), .1/

ci cj

n

o n o n o n o .1/ .1/ V ˝E.1/ C V ˝E.1/ ˝E.1/ PE 0 Veff ˝E.1/ PE ; 0 PE 0 Veff (2.77) where the hook represents a contraction. The first uncontracted term is represented by combinations of the diagrams in Fig. 2.1, such as ˝ .2/ PE D Q .E/

(2.78) considered as a single diagram. This diagram can be of two types.

28


?

?

?

?

Fig. 2.4 Examples of second-order wave-operator diagrams, excluding folded diagrams

If, on the one hand, both disconnected parts are open, the diagram is referred

to as linked.6 If, on the other hand, at least one of them is closed, the diagram is referred to as unlinked. .1/

In the unlinked part of the second term in (2.77), the closed part represents Veff , and since the order of the operators in the normal product is immaterial, this unlinked diagram appears also in the third term and is therefore eliminated. The last contracted term survives and represents the “folded” term. Here, the wave operator depends on the energy (E 0 ) of the intermediate state, which might differ from the energy of the initial state (E). We can then express the second-order wave operator by .1/ ˝ .2/ PE D Q .E/ V ˝E.1/ ˝E.1/ 0 PE 0 Veff

linked

PE ;

(2.79)

where only linked diagrams are maintained (see Fig. 2.4). The diagrams in Fig. 2.4 are second-order time-ordered Goldstone diagrams. In these diagrams, time is supposed to run from the bottom (although the formalism is here time independent). The diagrams are evaluated by the standard Goldstone rules with a denominator after each interaction equal to the energy difference between the (model-space) state at the bottom and that directly after the interaction (see Appendix I and [40, Sect. 12.4]). (In later chapters, we shall mainly use Feynman diagrams, which contain all possible time orderings between the interactions.)

2.4.2 Linked-Diagram Expansion 2.4.2.1 Complete Model Space Written more explicitly, the second-order wave operator (2.79) becomes

˝ .2/ PE D Q .E/ V Q .E/ V Q .E/ Q .E 0 / VPE 0 V linked PE :

6

(2.80)

A closed diagram has the initial as well as the final state in the model space. Such a diagram can – in the case of complete model space – have no other free lines than valence lines. A diagram that is not closed is said to be open.


29

Here, the second term has a double resolvent (double denominator, which might contain different model-space energies), and it is traditionally drawn in a “folded” way, as shown in the left diagram below (see, for instance, [40, Sect. 13.3])

r

r 6

s c a PE 0

)

d b

6s

c 6P 0 6d E a 6 6b PE

PE

(2.81)

The reason for drawing the diagram folded in this way is that the two pieces – before and after the fold – should be evaluated with their denominators independently. In the general case, by considering all possible time-orderings between the two pieces, together with the Goldstone evaluation rules, it can be shown that the denominators do factorize. In a relativistic treatment, which we shall employ for the rest of this book, the treatment is most conveniently based upon Feynman diagrams, which automatically contain all possible time-orderings, and then it is more natural to draw the diagram straight, as shown in the second diagram above. Factorization then follows directly. The double bar indicates that the diagram is “folded.” In such a diagram, the upper part has double denominators – one denominator with the energy of the initial state and one with that of the intermediate model-space state. The second-order wave operator can then be illustrated as shown in Fig. 2.5. Note that there is a minus sign associated with the folded diagram. The general ladder diagram (Fig. 2.5) may contain a (quasi)singularity, when the intermediate state lies in the model space and is (quasi)degenerate with the initial state. This singularity is automatically eliminated in the Bloch equation and leads to the folded term. Later, in Sect. 6.6 we shall discuss this kind of singularity in more detail in connection with energy-dependent interactions, and then we shall refer to the finite remainder as the model-space contribution (MSC). We have seen that the so-called unlinked diagrams are eliminated in the secondorder wave operator (2.79). When the model space is “complete” (see definition above), it can be shown that unlinked diagrams disappear in all orders of perturbation theory. This is the linked-cluster or linked-diagram theorem (LDE), first

r 6

6s

c 6 d P CQ 6 a 6 P

6b

)

r 6

6s

c 6Q

6d

a 6 P

6b

+

r 6

6s

c 6 P

6d

a 6 PE

6b

Fig. 2.5 Removing the singularity from a ladder diagram leads to finite remainder, represented by a “folded” diagram (last). The double bar represents a double denominator (with a factor of 1)

30


demonstrated in 1950s by Brueckner [9] and Goldstone [21] for a degenerate model space. It holds also for a complete quasi-degenerate model space, as was first shown by Brandow [7], using a double perturbation expansion. This was demonstrated more directly by Lindgren [37] by means of the generalized Bloch equation (2.56), and the result can then be formulated7 Œ˝; H0 P D .V ˝P ˝Veff /linked P:

(2.82)

This equation is a convenient basis for many-body perturbation theory, as developed, for instance, in [40]. It will also constitute a fundament of the theory developed in this book.

2.4.2.2 Incomplete Model Spaces When the model space is incomplete, i.e., does not contain all configurations that can be formed by the valence orbitals, the expansion is not necessarily completely linked. As first shown by Mukherjee [41, 56], the linked-diagram theorem can still be shown to hold, if the normalization condition (2.42a) is abandoned. As will be discussed later, a complete model space often has the disadvantage of so-called intruder states, which destroy the convergence. Then also other means of circumventing this problem will be briefly discussed.

2.5 All-Order Methods: Coupled-Cluster Approach 2.5.1 Pair Correlation Instead of solving the Bloch equation order-by-order, it is often more efficient to solve it iteratively. By separating the second-quantized wave operator into normalordered zero-, one-, two-,. . . body parts ˝ D ˝0 C ˝1 C ˝2 C

7

(2.83)

The Rayleigh–Schrödinger and the linked-diagram expansions have the advantage compared to, for instance, the Brillouin–Wigner expansion, that they are size-extensive, which implies that the energy of a system increases linearly with the size of the system. This idea was actually behind the discovery of the linked-diagram theorem by Brueckner [9], who found that the so-called unlinked diagrams have a nonphysical nonlinear energy dependence and therefore must be eliminated in the complete expansion. The concept of size extensivity should not be confused with the term size consistency, introduced by Pople [64, 65], which implies that the wave function separates correctly when a molecule dissociates. The Rayleigh–Schrödinger or linked-diagram expansions are generally not size consistent. The coupled-cluster approach (to be discussed below) does have this property in addition to the property of size extensivity.

2.5 All-Order Methods: Coupled-Cluster Approach

with

8 o n ˆ xji ˝ D c c ˆ 1 j i ˆ < o 1n ij ˝2 D ci cj cl ck xkl ; ˆ ˆ 2 ˆ : etc:

31

(2.84)

the Bloch equation can be separated into the following coupled n-particle equations Œ˝1 ; H0 P D .V ˝ ˝W /linked; 1 P; Œ˝2 ; H0 P D .V ˝ ˝W /linked; 2 P;

(2.85)

W D Veff D P V ˝P

(2.86)

etc., where is the effective interaction. Usually, the two-body operator dominates heavily, since it contains the important pair correlation between the electrons. Therefore, a good approximation for many cases is ˝ 1 C ˝1 C ˝2 ; (2.87) which yields Œ˝1 ; H0 P D .V1 C V ˝1 C V ˝2 ˝1 W1 /linked; 1 P; Œ˝2 ; H0 P D .V2 C V ˝1 C V ˝2 ˝1 W2 ˝2 W1 ˝2 W2 /linked; 2 P; (2.88) where W1 D .V1 C V1 ˝1 /closed; 1 ; W2 D .V2 C V ˝1 C V ˝2 /closed; 2 :

(2.89)

We see here that the equations are coupled, so that ˝1 appears in the equation of ˝2 and vice versa. This approach is known as the pair-correlation approach. Solving these coupled equations self-consistently is equivalent to a perturbation expansion – including one- and two-body effects – to essentially all orders. It should be noted, though, that each iteration does not correspond to a certain order of the perturbative expansion. As a simple illustration, we consider the simplified pair-correlation approach ˝ D ˝2

(2.90)

omitting single excitations. (This would be exact for a two-electron system using hydrogenic basis functions, in which case there are no core orbitals, but is a good approximation also in other cases.) The equation for ˝2 is Œ˝2 ; H0 P D .V C V ˝2 ˝2 W2 /linked; 2 P:

(2.91)

32


6

6

6

6

6

6

6

6

=

+

6

6

6

6+

6

6

6

6

6

6

6

6

6

6

+

+ folded

Fig. 2.6 Graphical representation of the pair function (2.96)

Operating on an initial two-electron state of energy E, the solution can be expressed ˝2 PE D Q .E/ .V C V ˝2 ˝2 W2 /linked PE :

(2.92)

Solving this iteratively leads to ˝2.1/ PE D Q .E/VPE ; .2/ .1/ .1/ .1/ PE ˝2 PE D Q .E/ V ˝2 ˝2 PE 0 W2

(2.93)

D Q .E/V Q .E/VPE Q .E/ Q .E 0 /VPE 0 VPE ; etc:;

(2.94)

where all terms are assumed to be linked. This leads to the “ladder sequence,” illustrated in Fig. 2.6. Note that in the expression above, all energies of the first term depend on the initial state, while in the folded term the wave operator depends on the energy of the intermediate state (E 0 ) (c.f., the “dot product,” introduced in Sect. 6.6). Operating with ˝2 in (2.84) on the initial statejabi leads to the pair function rs ˝2jabi D xab jrsi D ab .x1 ; x2 /;

(2.95)

which inserted in (2.88) leads to the pair equation ."a C "b h0 .1/ h0 .2// ab .x1 ; x2 / D .jrsi hrsjV jabi Cjrsi hrsjV jab i jcd i hcd jW2jabi/linked : (2.96) (For simplicity, we work with straight product functions – not antisymmetrized – in which case we sum over all combinations of r; s (without the factor of 1/2) with rs sr xab D xab .) We can also express the pair function as jab i D Q .E/ I Pairjabi ;

(2.97)


33

r r

s

r

= a

PE

b

+ a

s

s

PE

b

+

Q a

PE

b

r 6

6s

c 6P 0 E

6d

a

W2 b

PE

Fig. 2.7 Graphical representation of the self-consistent pair equation (2.99). The last diagram represents the “folded” term ˝2 W2 . The double line represents the double denominator (double resolvent)

where Q .E/ is the reduced resolvent (2.65) and E is the energy of the initial state jabi. I Pair represents the ladder sequence of Coulomb interactions (including folded terms), corresponding to the heavy line in Fig. 2.6, and including the resolvent (final denominator) leads to the pair function jab i. The effective interaction W2 can be expressed as (2.98) W2 D PE 0 I Pair PE ; which can be represented by the same diagrams as in Fig. 2.6 (with no final denominator) if the final state (with energy E 0 ) lies in the model space. The pair function (2.92) can now be expressed

Q .E/I Pair PE D Q .E/ V C V Q .E/I Pair Q .E 0 /I Pair PE 0 I Pair PE PE :

(2.99)

This relation can be represented graphically as shown in Fig. 2.7.

2.5.2 Exponential Ansatz A particulary effective form of the all-order approach is the Exponential Ansatz or Coupled-Cluster Approach (CCA), first developed in nuclear physics by Hubbard, Coster, and Kümmel [12, 13, 23, 33, 34]. It was introduced into quantum chemistry ˇ zek [11] and has been extensively used during the last decades for more deby Ciˇ tails. (The reader is referred to a recent book “Recent Progress in Coupled Cluster Methods” [79], which reviews the development of the methods since the start.) The CCA is a nonlinear approach, and the linear all-order approach (2.85), discussed above, is sometimes inadvertently referred to as “linear CCA”(!) – a term we shall not use here. In the exponential Ansatz, the wave operator is expressed in the form of an exponential 1 1 ˝ D eS D 1 C S C S 2 C S 3 C ; 2 3Š

(2.100)

34


S is the cluster operator (in chemical literature normally denoted by T ). It can then be shown that for a degenerate model space the cluster operator is represented by connected diagrams only.8 This implies that the linked but disconnected diagrams of the wave operator are here represented by the higher powers in the expansion of the exponential. For open-shell systems (with unfilled valence shell), it is convenient to represent the Ansatz in the normal-ordered form, introduced by Lindgren [38, 40], n o 1 ˚ 2 1 ˚ 3 S C S C : ˝ D eS D 1 C S C 2 3Š

(2.101)

This form has the advantage that unwanted contractions between the cluster operators are avoided. The cluster operator is completely connected also in this case, if the model space is complete [41], which can be formulated by means of the Bloch equation ŒS; H0 P D Q .V ˝P ˝Veff /conn :

(2.102)

Expanding the cluster operator in analogy with the wave-operator expansion (2.83) in terms on one-, two-,. . . body operators, S D S 1 C S2 C S3 C

(2.103)

yields n o 1 ˚ 3 1˚ 1˚ 1˚ S C : ˝ D eS D 1CS1 CS2 C S12 CfS1 S2 gC S22 C S12 S2 C 2 2 2 3Š 1 (2.104) With the approximation S D S1 C S 2 ;

(2.105)

the cluster operators satisfy the coupled Bloch equations ŒS1 ; H0 P D .V ˝ ˝W /conn; 1 P ŒS2 ; H0 P D .V ˝ ˝W /conn; 2 P

(2.106)

illustrated in analogy with Fig. 2.7 in Fig. 2.8. These equations lead to one- and twoparticle equations, analogous to the pair equation given above (2.96). Also these equations have to be solved iteratively, and we observe that they are coupled, as are the corresponding equations (2.88) for the full wave operator.

8 The distinction between linked and connected diagrams should be noted. A linked diagram can be disconnected, if all parts are open, as defined in Sect. 2.4.


6

S1 W

6

=

6

+

S2 W

6

=

6

6

6

6

6

6

6

+

6

6

6

+

6

+

6

6

+ +

6

6

6

6

6

6+

6

6

6

6

6

6

6

6

6

6

+

W2

W1

6

6

6

+

6

6

6

6

6

6

6

6

6

6

6

+

6

6

6

6

6

6

35

+

6

6

6

6

+

6

6

6

6

+

W1

6

6

6

6

6

6

6

6

6

6

+

+ +

W2

Fig. 2.8 Diagrammatic representation of the equations for the cluster operators S1 and S2 (2.106). The circle with a cross represents the “effective potential” in Fig. 2.3. The second diagram in the second row and the diagrams in the fourth row are examples of coupled-cluster diagrams. The last diagram in the second row and the three diagrams in the last row represent folded terms (c.f. Fig. 2.7)

The normal-ordered scheme is usually combined with a complete model space – or complete active space (CAS) – and the valence universality. This might lead to problems due to intruder states to be discussed further below. For atomic systems with essentially spherical symmetry, on the one hand, the cluster equations can be separated into angular and radial parts, where the former can be treated analytically and only the radial part has to have solved numerically

36


(see, for instance, [40, Chap. 15]). For molecular systems, on the other hand, analytical basis-set functions of Slater or Gaussian types are normally used to solve the coupled-cluster equations, as described in numerous articles in the field. As mentioned, the advantage of the normal ordering of the exponential Ansatz is that a number of unwanted contractions between open-shell operators is avoided. More recently, Mukherjee has shown that certain valence-shell contractions are actually desired, particularly when valence holes and strong relaxation are involved [26]. He then introduced a modified normal ordering ˝ D ffexp.S /gg;

(2.107)

where contractions involving passive (spectator) valence lines are reintroduced compared to the original normal ordering.

2.5.3 Various Models for Coupled-Cluster Calculations: Intruder-State Problem The early forms of coupled-cluster models were of single-reference type (SRCC) with a one-dimensional (closed-shell) model space. In the last few decades, various versions of multireference (MRCC) models with multidimensional model space have appeared (for reviews, see e.g., [2, 41, 58]). These are essentially two major types, known as valence-universal multireference (VU-MRCC) [38, 57] and state-universal multireference (SU-MRCC) [27, 28] methods, respectively. In the valence-universal methods, the same cluster operators are being used for different ionization states and therefore particularly useful for calculating ionization energies and affinities. In the state-universal methods, specific operators are used for a particular ionization stage and particularly used when different states of the same ionization are considered or in the molecular case for studying potential energy surfaces (PES). A serious problem that can appear in MBPT with a multireference model space is what is known as the “intruder-state-problem.” This appears when a state outside the model space – of the same symmetry as the state under consideration – has a perturbed energy between those of the same symmetry originating from the model space. This will destroy the convergence of the perturbation expansion. This problem was first observed in nuclear physics [74], but it was early observed also in atomic physics for the beryllium atom [69]. Here, the ground state is 1s 2 2s 2 1 S , and the excited state 1s 2 2p 2 1 S has a low unperturbed energy, while the true state lies close to the 2s ionization limit. This implies that when the perturbation is gradually turned on, a large number of “outside” states, 1s 2 2s ns, will cross the energy of the 1s 2 2p 2 1 S state, and there will be no convergence beyond the crossing point. The convergence problem due to intruders is particularly serious in perturbation theory, when the states are expanded order-by-order from the unperturbed ones. In the coupled-cluster approach, which in principle is nonperturbative, it might be

2.6 Relativistic MBPT: No-Virtual-Pair Approximation

37

possible to find a self-consistent solution of the coupled equations without reference to any perturbative expansion. It was first shown by Jankowski and Malinowski [24, 25, 48] that it was in fact possible to find a solution to the beryllium problem with a complete model space. Lindroth and M˚artensson [44] solved the same problem by means of complex rotation. Several other methods have been developed to reduce the intruder-state problem. One way is to reduce the model space and make it incomplete. It was shown by Mukherjee [56] that by abandoning the intermediate normalization (2.42a), the linked character of the diagram expansion could still be maintained. The criteria for the connectivity of the coupled-cluster expansion have been analyzed by Lindgren and Mukherjee [41]. Another approach to avoid or reduce the intruder-state problem is to apply an intermediate-effective Hamiltonian, a procedure developed by the Toulouse group (Malrieu, Durand et al.) in the mid 1980s [15]. Here, only a limited number of roots of the secular equation are being looked for. A modified approach of the method has been developed by Meissner and Malinowski [51] and applied to the abovementioned beryllium case. A third approach to the problem is the state-specific multireference (SS-MRCC) approach, where a multireference is used but only a single state is considered [47]. This approach can be regarded as an extreme of the intermediate-Hamiltonian approach and is frequently used particularly for studying potential-energy surfaces. All the coupled-cluster approaches can also be applied in the relativistic formalism, although applications are here still quite limited. We shall return briefly to this problem in Chap. 8.

2.6 Relativistic MBPT: No-Virtual-Pair Approximation In setting up a Hamiltonian for relativistic quantum mechanics, it may be tempting to replace the single-electron Schrödinger Hamiltonian in the many-body Hamiltonian (2.11) by the Dirac Hamiltonian (see Appendix D) hD D c˛ pO C ˇmc 2 C vext ;

(2.108)

which with the Coulomb interaction between the electrons VC D

N X i <j

e2 40 rij

(2.109)

yields the Dirac–Coulomb Hamiltonian HDC D

N X i D1

hD .i / C VC :

(2.110)

38


This Hamiltonian, however, has several serious shortcomings. First, it is not bound from below, because nothing prevents the electrons from falling into the “Dirac sea” of negative-energy electron states. A many-electron state with a mixture of negative-energy and positive-energy electron states can then be accidentally degenerate with a state with only positive-energy states – a phenomenon known as the Brown–Ravenhall disease [8]. In Chap. 6, we shall derive a field-theoretical manybody Hamiltonian that will be used in the further development. In this model, there is no “Dirac sea,” but the negative-energy states correspond to the creation of positron states, which are highly excited. Then there can be no Brown–Ravenhall effect. Within the conventional many-body treatment, the Brown–Ravenhall effect can be circumvented by means of projection operators [78], which exclude negativeenergy states, leading to the projected Dirac–Coulomb Hamiltonian " HDCproj D C

N X

# hD .i / C VC C :

(2.111)

i D1

Including also the instantaneous Breit interaction (see Appendix F) " # e 2 X ˛i ˛j .˛i r ij /.˛j r ij / C VB D ; 80 r ij rij2

(2.112)

i t0 GC .x; x0 / D .t t0 / G.x; x0 /;

(5.2)

where .t/ is the Heaviside step function (Appendix A.29), which implies Z .t t0 / .x/ D

d3 x 0 GC .x; x0 / .x0 /:

(5.3)

We assume now that the function .x/ satisfies a differential equation of Schrödinger type @ (5.4) i H.x/ .x/ D 0: @t


91

92

5 Green’s Functions

Operating with the bracket on (5.3) yields @ d x 0 i H.x/ GC .x; x0 / .x0 / @t

Z

3

iı.t t0 / .x/ D

using the relation (A.31), which implies that the retarded Green’s function satisfies the differential equation

@ i H.x/ GC .x; x0 / D iı 4 .x x0 /: @t

(5.5)

– A relation often taken as the definition of the (mathematical) Green’s function. The Green’s function can be used for solving inhomogeneous differential equations. If we have @ i H.x/ .x/ D f .x/; (5.6) @t then the solution can be expressed Z .x/ D

dx 0 GC .x 0 ; x/ f .x 0 /:

(5.7)

5.2 Field-Theoretical Green’s Function: Closed-Shell Case 5.2.1 Definition of the Field-Theoretical Green’s Function In the closed-shell case, the field-theoretical single-particle Green’s function can

be defined [5]1 G.x; x0 / D

iˇ E D ˇ h ˇ ˇ 0H ˇT O H .x/ O H .x0 / ˇ 0H h0H j 0H i

;

(5.8)

where T is the Wick time-ordering operator (2.27) and O H ; O H are the electronfield operators in the Heisenberg representation (HP) (B.27). The statej0H i is the “vacuum in the Heisenberg representation,” i.e., the state in the Heisenberg representation with no particles or holes. In a “closed-shell state,” the single reference or model state is identical to the vacuum state (see Sect. 2.3). 1

Different definitions of the field-theoretical Green’s function are used in the literature. The definition used here agrees with that of Itzykson and Zuber [7], while that of Fetter and Walecka [5] differs by a factor of i.

5.2 Field-Theoretical Green’s Function: Closed-Shell Case

93

The Heisenberg vacuum is time independent and equal to the corresponding vacuum state in the interaction picture at t D 0, i.e., j0H i D U.0; 1/j0i ;

(5.9)

where U.t; t0 / is the evolution operator (3.6) andj0i is the unperturbed vacuum or the IP vacuum as t ! 1 [c.f. (3.28)]. Using the relation between the electron-field operators in HP and IP (B.25) O H .x/ D U.0; t/ O .x/ U.t; 0/;

(5.10)

we can transform the Green’s function (5.8) to the interaction picture G.x; x0 / D

ˇ E i h D ˇ ˇ ˇ 0 ˇU.1; 0/T U.0; t / O .x/U.t; 0/U.0; t0 / O .x0 /U.t0 ; 0/ U.0; 1/ˇ 0 h0jU.1; 1/j0i

:

(5.11)

For t > t0 , the numerator becomes ˇ E D ˇ ˇ ˇ 0 ˇU.1; t/ O .x/U.t; t0 / O .x0 /U.t0 ; 1/ˇ 0 and for t < t0

ˇ E D ˇ ˇ ˇ 0 ˇU.1; t0 / O .x0 /U.t0 ; t/ O .x/U.t; 1/ˇ 0

(5.12a)

(5.12b)

using the relation (3.8). From the expansion (3.15), we obtain the identity Z t Z 1 X .i/ t dt1 : : : dt T V .t1 / : : : V .t / e.jt1 jCjt2 j / U.t; t0 / D Š t0 t0 D0 Z Z 1 t X .i/n t D dt1 : : : dtn T V .t1 / : : : V .tn / e.jt1 jCjt2 j / nŠ t1 t1 nD0 Z t1 Z 1 t m X .i/ 1 dt1 : : : dtm T V .t1 / : : : V .tm / e.jt1 jCjt2 j / ; mŠ t0 t0 mD0 (5.13) where we have included the unity as the zeroth-order term in the summation. If we concentrate on the :th term of the first sum, we have the identity (leaving out the damping factor) 1 Š

Z

t0

D

Z

t

dt1 : : : X

mCnD

t

t0

dt T V .t1 / : : : V .t /

1 mŠ nŠ

Z

Z

t t1

dt1 : : :

t

t1

dtn T

Z

Z

t1 t0

dt1 : : :

t1

t0

dtm T : (5.14)

94


We can now apply this identity to the first part of the numerator (5.12a), U.1; t/ O .x/U.t; t0 /. The interaction times of U.1; t/, O .x/ and U.t; t0 / are time ordered, and hence the result can be expressed Z 1 Z h i 1 1 dt1 : : : dt T V .t1 / : : : V .t / O .x/ : (5.15) Š t0 t0 The same procedure can be applied to the rest of the expression (5.12a) as well as to the other time ordering (5.12b). With the perturbation (3.16), the numerator of the single-particle Green’s function (5.11) then becomes [5, Eq. 8.9] Z Z 1 D ˇ h iˇ E X 1 i n ˇ ˇ d4 x1 d4 xn 0H ˇT O H .x/ O H .x0 / ˇ 0H D nŠ c nD0 D ˇ h iˇ E ˇ ˇ 0 ˇT O .x/ H.x1 / H.xn / O .x0 / ˇ 0 e.jt1 jCjt2 j / (5.16) with integrations over all internal times. In transforming the time-ordering to normal ordering by means of Wick’s theorem, only fully connected terms remain, since the vacuum expectation of any normal-ordered expression vanishes (see Sect. 4.41). The denominator in (5.8) becomes, using the relation (5.9), h0H j 0H i D h0jU.1; 1/j0i D h0jS j0i; where S is the S-matrix (4.2). Then the Green’s function can be expressed

D G.x; x0 / D

ˇ h iˇ E ˇ ˇ 0H ˇT O H .x/ O H .x0 / ˇ 0H h0jS j0i

:

(5.17)

We see that this expansion is very similar to that of the S-matrix (4.3), the main difference being the two additional electron-field operators. Therefore, the Green’s function can also be expressed as

G.x; x0 / D

D ˇ h iˇ E ˇ ˇ 0 ˇT O .x/U.1; 1/ O .x0 / ˇ 0 h0jS j0i

;

(5.18)

where the time-ordered product is connected to form a one-body operator. This leads to Z Z 1 X 1 i n 1 G.x; x0 / D d4 x1 d4 xn h0jSj0i nD0 nŠ c D ˇ h iˇ E ˇ ˇ 0 ˇT O .x/ H.x1 / H.xn / O .x0 / ˇ 0 e.jt1 jCjt2 j / : (5.19) The Green’s function is like the S-matrix Lorentz covariant.


95

The two-particle Green’s function is defined in an analogous way iˇ ˛ ˝ ˇ h 0ˇT O H .x/ O H .x 0 / O H .x00 / O H .x0 / ˇ0 G.x; x 0 I x0 ; x00 / D h0jS j0i

(5.20)

and transforming to the interaction picture leads similarly to Z Z 1 X 1 i n 1 G.x; x 0 I x0 ; x00 / D d4 x1 d4 xn h0jSj0i nD0 nŠ c h0jT O .x/ O .x 0 / H.x1 / H.xn / O .x00 / O .x0 / j0i e.jt1 jCjt2 j / (5.21) and analogously in the general many-particle case. Note that the coordinates are here four-dimensional space-time coordinates, which implies that the particles have individual initial and final times. This is in contrast to the quantum-mechanical wave function or state vector, which has the same time for all particles. We shall discuss this question further below. We can transform the time-ordered products above to normal-ordered ones by means of Wick’s theorem (see Sect. 2.2.3). Since normal-ordered products do not contribute to the vacuum expectation value, it follows that only fully contracted contribute to the Green’s function. The contractions between the electron-field operators and the interaction operators lead to electron propagators (SF ) (4.9) on the in- and outgoing lines as well as all internal lines (see Fig. 5.1).2 This allows time to run in both directions, and both particle and hole states can be involved.

5.2.2 Single-Photon Exchange The Green’s function for single-photon exchange in Fig. 5.2 can be constructed in close analogy to that of the corresponding S-matrix in Sect. 4.4, ZZ G.x; x 0 ; x0 ; x00 / D d4 x2 d4 x1 iSF .x; x1 / iSF .x 0 ; x2 / .i/e 2 DF .x2 ; x1 / iSF .x1 ; x0 / iSF .x2 ; x00 / e.jt1 jCjt2 j/ : (5.22) x

x

x0

SF

SF

SF

SF

SF

SF

x0

x0

x00

Fig. 5.1 Graphical representation of the one- and two-particle Green’s function. The orbital lines between dots represent electron propagators 2

In our notations, an orbital line between heavy dots always represents an electron propagator.

96

5 Green’s Functions x0

x

Fig. 5.2 Green’s function for single-photon exchange

SF

!3

!4 SF z

1 SF x0

!1 E0

2 !2 SF x00

With the transforms (4.10) and (4.31), this becomes after integrating over the internal limes [using the relation (A.17)] ZZ ZZ d!3 d!4 0 0 G.x; x 0 ; x0 ; x00 / D eit !3 eit !4 eit0 !1 eit0 !2 d3 x 1 d 3 x 2 2 2 Z ZZ dz d!1 d!2 iSF .!3 I x; x 1 / 2 2 2 iSF .!4 I x 0 ; x 2 / .i/e 2 DF .zI x 2 ; x 1 / iSF .!1 I x 1 ; x 0 / iSF .!2 I x 2 ; x 00 /2 .!1 z !3 / 2 .!2 C z !4 /: (5.23) In the equal-time approximation, where the particles have the same initial and final times (t D t 0 and t0 D t00 ), the external time dependence becomes eit .!3 C!4 / eit0 .!1 C!2 / . In the limit ! 0, we have after z-integration !1 C !2 D !3 C !4 , and if we consider the diagram as a part of a ladder, this is equal to the initial energy E0 . We define the Feynman amplitude for the Green’s function as the function with the external time dependence removed. This gives G.x; x 0 ; x0 ; x00 / D Msp .x; x 0 I x 0 ; x 00 / ei.t t0 /E0 and Msp .x; x 0 I x 0 ; x 00 / D

(5.24)

ZZ d!1 d!2 d!3 d!4 2 2 2 2 iSF .!4 I x 0 ; x 2 /.i/I.!1 !3 I x 2 ; x 1 / iSF .!1 I x 1 ; x 0 /

ZZ

d3 x 1 d 3 x 2

ZZ

iSF .!2 I x 2 ; x 00 /22 .!1 C !2 !3 !4 /

(5.25)

using the definition (4.44).

5.2.3 Fourier Transform of the Green’s Function 5.2.3.1 Single-Particle Green’s Function Assuming the Heisenberg vacuum state j0H i to be normalized, the single-particle Green’s function (5.8) becomes


97

˝ ˇ ˇ ˛ G.x; x0 / D 0H ˇT O H .x/ O H .x0 / ˇ0H ˇ ˛ ˇ ˛ ˝ ˇ ˝ ˇ D .t t0 / 0H ˇ O H .x/ O H .x0 /ˇ0H .t0 t/ 0H ˇ O H .x0 / O H .x/ˇ0H : (5.26) The retarded part (5.2) is then, using the relation (B.27) in Appendix B, ˇ ˛ ˝ ˇ GC .x; x0 / D 0H ˇ O H .x/ O H .x0 /ˇ0H ˇ ˛ ˝ ˇ D 0H ˇ eiHt O S .x/ eiHt eiHt0 O S .x 0 / eiHt0 ˇ0H :

(5.27)

Inserting between the field operators a complete set of positive-energy eigenstates of the second-quantized Hamiltonian H (2.17), corresponding to the .N C 1/-particle system H jni D En jni (5.28) yields the Lehmann representation GC .x; x0 / D

X˝

ˇ ˛ ˇ ˇ ˛ ˝ ˇ 0H ˇeiHt O S .x/ˇn eiEn .t t0 / nˇ O S .x0 / eiHt ˇ0H

(5.29)

n

summed over the intermediate states of the .N C 1/ system. The ground state as well as the inserted intermediate states are eigenstates of the Hamiltonian H , and setting the energy of the former to zero, this yields GC .x; x0 / D

X˝

ˇ ˛ ˇ ˛ ˇ ˝ ˇ 0H ˇ O S .x/ˇn eiEn .t t0 / nˇ O S .x0 /ˇ0H :

(5.30)

n

Performing a Fourier transform of the Green’s function, including the adiabatic damping e (see Sect. 3.2), yields ( D t t0 > 0) Z GC .EI x; x 0 / D using

1

iE

d e 0

Z

1

ˇ ˛˝ ˇ ˇ ˛ ˝ ˇ 0H ˇ O S .x/ˇn nˇ O S .x 0 /ˇ0H GC . ; x; x 0 / D i E En C i (5.31)

dt ei˛t e t D

0

i : ˛ C i

(5.32)

Analogous results are obtained for the advanced part (t < t0 ) of the Green’s function, corresponding toˇ a holeˇ in the initial system. ˝ ˛ The expression nˇ O S .x/ˇ0H represents a state n .x/ of the .N C 1/ system in the Schrödinger picture, and an equivalent expression of the Fourier transform of the Green’s function becomes

GC .EI x; x 0 / D i

X n .x/ .x 0 / n

n

E En C i

:

(5.33)

98


This implies that the poles of the retarded/advanced single-particle Green’s function represent

the true energies of the vacuum plus/minus one particle, relative to the vacuum state. In order to show that the definition (5.8) of the Green’s function is compatible with the classical definition [(5.1) and (5.5)], we form the reverse transformation Z GC .x; x0 / D GC . ; x; x 0 / D

dE iE X n .x/n .x 0 / e : i 2 E En C i n

(5.34)

We then find that Z @ dE iE X E En i H.x/ GC .x; x0 / D i e n .x/ n .x 0 /: @t 2 E E C i n n (5.35) Letting ! 0 and using the closure property (C.27) X n

n .x/ n .x 0 / D ı 3 .x x 0 /

and the integral Z

dE iE e D ı. / D ı.t t0 /; 2

we confirm that the retarded part of the Green’s function (5.8) satisfies the relation (5.5) @ (5.36) i H.x/ GC .x; x0 / D iı 4 .x x0 /: @t

5.2.3.2 Electron Propagator We consider now the zeroth-order single-particle Green’s function (5.19) ˇ E D ˇ ˇ ˇ G0 .x; x0 / D 0 ˇT Œ O .x/ O .x0 /ˇ 0 ;

(5.37)

where the vacuum and the field operators are expressed in the interaction picture. Then we find that the single-particle Green’s function is identical to the Feynman electron prop-

agator (4.9) times the imaginary unit i G0 .x; x0 / iSF .x; x0 /:

(5.38)


99

The retarded operator can be transformed in analogy with the Lehmann representation above G0C .x; x0 / D

ˇ ˇ E ˇ X D ˇˇ 0 ˇ ˇ ˇ 0H ˇ O S .x/ˇ n0 eiEn .t t0 / n0 ˇ O S .x0 /ˇ 0H ;

(5.39)

n

˛ wherejn0 are eigenstates of the zeroth-order Hamiltonian for the .N C 1/-particle system (B.22) ˛ ˛ H0jn0 D En0jn0 and En0 are the energies relative the vacuum. Performing the time integration yields the Fourier transform ˛˝ X hxjn0 n0 jx 0 i G0C .x; x 0 ; E/ D i : (5.40) E En0 C i n The corresponding advanced function becomes ˛˝ X hxjn0 n0 jx 0 i : G0 .x; x 0 ; E/ D i E En0 i n

(5.41)

Both of these results can be expressed by means of a complex integral Z G0 .x; x0 / D iSF .x; x0 / D i

˛˝ dE hxjn0 n0 jx 0 i iE.t t0 / e ; 2 E En0 C i n

(5.42)

where n has the same sign as En0 , i.e., positive for particle states and negative for hole or antiparticle states. The zeroth-order Green’s function or electron propagator can also be expressed in operator form as GO 0 .E/ D iSOF .E/ D

i : E H0 ˙ i

(5.43)

5.2.3.3 Two-Particle Green’s Function in the Equal-Time Approximation Setting the initial and final times equal for the two particles, t D t 0 and t0 D t00 , the retarded two-particle Green’s function (5.20) becomes ˇ ˛ ˝ ˇ GC .x; x 0 I x0 ; x00 / D 0H ˇ O H .x/ O H .x 0 / O H .x00 / O H .x0 /ˇ0H ˝ ˇ D 0H ˇ eiHt O S .x/ O S .x 0 /ei Ht ˇ ˛ eiHt0 O S .x 00 / O S .x 0 /eiHt0 ˇ0H :

(5.44)

100


We introduce a complete set of two-particle states (5.28), which leads to the Lehmann representation ˇ E ˇ E D ˇ X D ˇˇ ˇ ˇ ˇ GC .x; x 0 I x0 ; x00 / D 0H ˇ O S .x/ O S .x 0 /ˇ n eiEn .t t0 / n ˇ O S .x00 / O S .x0 /ˇ 0H n

(5.45)

with the Fourier transform

ˇ ˛˝ ˇ ˇ ˛ ˝ ˇ X 0H ˇ O S .x/ O S .x 0 /ˇn nˇ O .x 00 ; O .x 0 //ˇ0H S S D GC .EI x; x E En ˙ i n (5.46) ˝ ˇ with the upper (lower) sign for the retarded (advanced) function. Here, nˇ O S .x 0 / ˇ ˛ O .x 00 /ˇ0H represents a two-particle state n .x; x 0 / in the Schrödinger picture, S which yields the Fourier transform 0

I x 0 ; x 00 /

GC .EI x; x 0 I x 0 ; x 00 / D i

X n .x; x 0 / .x 0 ; x 0 / n

E En ˙ i

n

0

:

(5.47)

This implies that also in this case the poles of the Green’s function represent the exact eigenvalues of the system, relative to the vacuum. Note that this holds in the many-particle case only in the equal-time approximation, where there is only a single time coordinate D t t0 .

5.3 Graphical Representation of the Green’s Function We shall now demonstrate how the expansions of the Green’s-functions [(5.19) and (5.21)] can be conveniently represented by means of Feynman diagrams [6], discussed in the previous chapter, and we start with the single-particle case.

5.3.1 Single-Particle Green’s Function The zeroth-order Green’s function is (with our definition) identical to the Feynman electron propagator times the imaginary unit (5.38) or equal to the contraction (4.9) D ˇ h iˇ E ˇ ˇ G0 .x; x0 / D 0 ˇT O .x/ O .x0 / ˇ 0 D O .x/ O .x0 /; (5.48) which we represent graphically as in Fig. 4.1. x

G0 .x; x0 / D

6j x0

(5.49)

5.3 Graphical Representation of the Green’s Function

101

This contains both time orderings, i.e., j represents both particle and hole/ antiparticle states. In next order, the numerator of the Green’s function (5.16) has the form ZZ ˇˇ E 1 D ˇˇ (5.50) 2 0ˇ d4 x1 d4 x2 T O .x/ H.x1 /H.x2 / O .x0 / ˇ0 : 2c The photon fields have to be contracted, which leads to a two-particle interaction, in analogy with the single-photon interaction Vsp (4.44), H.x1 /H.x2 / D v.x1 ; x2 /

(5.51)

with the Fourier transform with respect to time v.zI x 1 ; x 2 /, which we represent graphically as 1

2 (5.52)

The vacuum expectation (5.50) can then be illustrated by the following picture x

h0jTŒ

1

2

]j0i

x0

(5.53)

where the vertical lines represent the electron-field operators. The procedure is now to transform the time ordering to normal ordering (see Sect. 2.2), which we can do by means of Wick’s theorem (2.34). This leads to a normal-ordered totally uncontracted and all possible normal-ordered single, doubly, . . . contracted terms. In the vacuum expectation only fully contracted terms will survive. We can here distinguish between two cases: either the electron-field operators are connected to each other and disconnected from the interaction or all parts are connected to a single piece. The former case leads to the diagrams

6 6

-

6

6

-

(5.54)

102


where the disconnected, closed parts represent the closed first-order S-matrix diagrams Scl.1/

D 6

-

6C

:-

(5.55)

The diagrams in (5.54) can then be expressed G0 Scl.1/ . Connecting all parts of the expression (5.53) leads to the diagrams

-

(5.56) These diagrams are quite analogous to the S-matrix diagrams for vacuum polarization and self-energy, discussed in Sect. 4.6, the only difference being that the Green’s-function diagrams contain in- and outgoing electron propagators. We note that all internal lines do represent electron propagators, containing particle as well as hole states. We can now see that the disconnected parts of the diagrams (5.54) are eliminated by the denominator in the definition of the Green’s function (5.8). Therefore, we can then represent the Green’s function up to first order by connected diagrams only (Fig. 5.57).

6

+

+

(5.57)

We shall now indicate that this holds also in higher orders. Next, we consider a Green’s-function term with two two-particle interactions.

h0jT [

]j0i (5.58)

We can here distinguish different cases.


103

We consider first the case where both interactions are disconnected from the electron-field operators. Leaving out the latter, we then have

h0jT[

j0i (5.59)

This corresponds to the vacuum expectation of the second-order S-matrix and leads to connected diagrams 6 -

6 ? 6 ? -

6

6

-

6 (5.60)

and to the disconnected diagrams 6

-

66

-

6

6

-

6-

-

(5.61)

.2/

We denote these diagrams by Scl D h0jS .2/j0i. In addition, we have the free electron-field operators, which combine to the zeroth-order Green’s function G0 . Therefore, we can express the corresponding GF diagrams as G0 Scl.2/ . Next, we consider the case where one of the interactions in (5.59) is closed by itself, while the remaining part is connected. This leads to disconnected diagrams, where the disconnected part is the closed first order (5.55) and the connected part is identical to the connected first-order diagrams in Fig. 5.56, which we can express the disconnected diagram as GC.1/ Scl.1/ . Finally, we have the case where all diagram parts are completely connected, shown in Fig. 5.3, which we denote by GC.2/ . Going to third order, we find similarly that we can have G0 D G .0/ combined with the closed diagrams Scl.3/ , GC.1/ combined with Scl.2/ , GC.2/ combined with Scl.1/ .3/ and finally completely connected GC diagrams. This leads to the sequence 8 .0/ G ˆ ˆ ˆ .1/ ˆ .0/ .1/ ˆ < GC C G Scl .2/ .1/ .1/ .2/ ; GC C GC Scl C G .0/ Scl ˆ ˆ .3/ .2/ .1/ .1/ .2/ .0/ .3/ ˆ G C G S C G S C G S ˆ C cl C cl cl ˆ : C etc:; which summarizes to .G .0/ C GC.1/ C GC.2/ C /.1 C Scl.1/ C Scl.2/ C / D .G0 C GC /.1 C Scl/; (5.62)

104


-

6

-

6 ?

6

-

6

Fig. 5.3 Second-order connected diagrams of the one-body Green’s function, assuming a twobody interaction

where GC represents all connected diagrams of the numerator of the GF expression (5.21) and Scl represents all closed S diagrams. But the last factor is the vacuum expectation of the S matrix to all orders h0jSj0i D 1 C Scl ;

(5.63)

which implies that this is canceled by the denominator in the definition (5.8). Hence, the single-particle Green’s function can in the close-shell case be represented

by completely connected diagrams " 1 Z X 1 1 n Z d4 x1 d4 x2n iG.x; x0 / D nŠ c 2 nD0

D E 0T O .x/ H.x1 ; x2 / H.x2n1 ; x2n / O .x0 / 0

# : conn

(5.64) This can also be expressed ˇ ˛ ˝ ˇ G.x; x0 / D 0H ˇT Œ O H .x/ O H .x0 /ˇ0H conn :

(5.65)

The connectedness of the Green’s function can also be shown in a somewhat different way. If we remove the two electron-fields operators and the denominator from the Green’s function expansion (5.19), then we retrieve the vacuum expectation of


105

the S-matrix (4.3) h0jS j0i. Therefore, if the field operators are connected to each other and the interactions among themselves, the result (after including the denominator) is simply the zeroth-order Green’s function iG .0/ . If the field operators are connected to one of the interactions, they form the connected first-order Green’s .1/ function iGconn and the remaining interactions again form h0jS j0i. Continuing the process leads to .1/ .2/ G D G .0/ C Gconn C Gconn C ; (5.66) which proves that the single-particle Green’s function is entirely connected. 5.3.1.1 One-Body Interaction We shall now consider the case when we, in addition to the two-body interaction, have a one-body interaction of potential type

(5.67)

The graphical representation can then be constructed in the same way as before, and we then find in first order the additional diagrams

66

-

(5.68) The first diagram is unconnected, and the closed part is a part of h0jS j0i and hence this diagram is eliminated by the denominator of (5.19), as before. It is not difficult to show that the single-particle Green’s function is represented by connected diagrams only, when we have a mixture of one- and two-body interactions. The additional connected diagrams in second order are shown in Fig. 5.4.

5.3.2 Many-Particle Green’s Function We now turn to the two-particle Green’s function (5.21). The zeroth-order Green’s function is in analogy with the one-particle function (5.49) represented by x G0 .x; x 0 I x0 ; x00 / D

x0 0 0 6 6 D iSF .x; x0 / iSF .x ; x0 /

x0

x00

or a product of two Feynman electron propagators.

(5.69)

106


-

6

6

-

-

6

Fig. 5.4 Additional second-order diagrams of the single-particle Green’s function – in addition to those in Fig. 5.3 – with a combination of one- and two-body interactions

As mentioned earlier, the (initial and final) times of the two particles in principle can be different, although we shall in most applications assume that they are equal, as will be further discussed in the following. In first order, we have in analogy with the single-particle case (5.53)

h0jT Œ

]j0i (5.70)

This can lead to disconnected diagrams, composed of the zeroth-order function (5.69) and the closed first-order diagrams (5.55). Another type of disconnected diagrams is the combination of zeroth-order single-particle GF and the connected first-order GF

6

6

(5.71)

It should be noted that both parts are here consider as open (not closed).3 Finally, we can have an open two-particle diagram

3

Generally, a diagram is considered closed if it has no free lines/propagators, like the diagrams in (5.60) and (5.61), while an open diagram has at least one pair of free lines, like those in Fig. 5.3. An operator or a function represented by a closed/open diagram is said to be closed/open.


107

(5.72) In second order, we can have the zeroth-order two-particle Green’s function, combined with second-order closed diagrams, Scl.2/ , and connected first-order diagrams combined with first-order closed diagrams, Scl.1/ . In addition, we can have disconnected diagrams with two open first-order single-particle diagrams (5.56). Continuing this process leads formally to the same result as in the single-particle case (5.62) – the diagrams with a disconnected closed part are eliminated by the denominator. Formally, the diagrams can still be disconnected, like (5.71), since there is a disconnected zeroth-order Green’s function part. We shall refer to such diagrams as linked in analogy with the situation in MBPT (Sect. 2.4). The result is then expressed "

G.x; x

0

I x0 ; x00 /

Z Z 1 D ˇ h X 1 1 n ˇ 4 4 d D x d x 0 ˇT O .x/ O .x 0 / 1 2n 2 nŠ c nD0 # i ˇE ˇ v.x1 ; x2 / v.x2n1 ; x2n / O .x00 / O .x0 / 0ˇ ; (5.73) linked

a result that can easily be extended to the general many-particle case. The two-body interactions used here correspond to two contracted interactions of the type (4.4). Uncontracted interactions of this kind cannot contribute to the Green’s function, since this is a vacuum expectation. Therefore, the results above can in the single-particle case also be expressed Z Z 1 X 1 i n 4 G.x; x0 / D d x1 d4 xn nŠ c nD0 E D ˇ ˇˇ ˇ 0 ˇT O .x/ H.x1 / H.xn / O .x0 / ˇ 0conn

(5.74)

including even- as well as odd order terms, and similarly in the many-particle case. This can also be expressed ˇ ˛ ˝ ˇ G.x; x0 / D 0H ˇT Œ O H .x/ O H .x0 /ˇ0H conn

(5.75)

and in the two-particle case ˇ ˛ ˝ ˇ G.x; x 0 I x0 ; x00 / D 0H ˇT Œ O H .x/ O H .x 0 / O H .x00 / O H .x0 /ˇ0H linked :

(5.76)

108


The linked character of the Green’s function can also in the two-particle case be shown as we did at the end on the single-particle section. If all interactions of the expansion (5.21) are connected among themselves, they form the vacuum expectation value of the S-matrix, canceling the denominator, and the electron-field operators .0/ form the two-body zeroth-order Green’s function G2 . If one pair of field operators are internally connected, then the remaining part is identical to the single-particle Green’s function G1 , which has been shown to be connected. The result G1.0/ G1 is disconnected but since both parts are open, this is linked with the convention we use. If one pair of field operators are connected to some of the interactions and the other pair to the remaining ones, the result is G1 G1 , which is also disconnected but linked. Finally, all field operators can be connected to the interactions, which leads to the connected two-particle Green’s function G2;conn . The remaining interactions form h0jSj0i, canceling the denominator, and the result becomes G2;conn . In summary, the two-particle Green’s function becomes G2 D G2.0/ C G1;conn G1;conn C G2;conn ;

(5.77)

which can be disconnected but linked. This argument can easily be generalized, implying that the many-particle Green’s function in the closed-shell case is linked.

5.3.3 Self-Energy: Dyson Equation All diagrams of the one-particle Green’s function can be expressed in the form ZZ G.x; x0 / D G.x; x0 / C d4 dx1 d4 dx2 G0 .x; x1 / .i/˙.x1 ; x2 / G0 .x2 ; x0 /; (5.78) where ˙.x2 ; x1 / represents the self-energy. This can be represented as shown in Fig. 5.5, i.e., as the zeroth-order Green’s function plus all self-energy diagrams. Some of the second-order self-energy diagrams in Figs. 5.3 and 5.4 have the form of two first-order diagrams, connected by a zeroth-order GF. All diagrams of that kind can be represented as a sequence of proper self-energy diagrams, ˙ , which have the property that they cannot be separated into lower-order diagrams by cutting a single line. This leads to the expansion of the total self-energy shown in Fig. 5.6, where the crossed box represents the proper self-energy. The single-particle Green’s function can then be represented as shown in Fig. 5.7, which corresponds to the Dyson equation for the single-particle Green’s function. ZZ G.x; x0 / D G0 .x; x0 / C

d4 x1 d4 x2 G0 .x; x2 / .i/˙ .x2 ; x1 / G.x1 ; x0 /: (5.79)


109

Fig. 5.5 The single-particle Green’s function expressed in terms of the self-energy

6 6

=

6

6+

6

Fig. 5.6 Expansion of the total self-energy in terms of proper self-energies. The crossed box represents the proper self-energy ˙

6

=

6

x0

+

6

6 6

=

6+ 6 6

x0

x

x00

6

6

= x0

+

6

Fig. 5.7 Graphical representation of the Dyson equation for the single-particle Green’s function (5.79), using the proper self-energy ˙

x

6

6

6

6

+

x00

x0

x

x0

x2

x20

x1

x10

˙ .x2 ; x20 I x1 ; x10 / G.x1 ; x10 I x0 ; x00 /

x00

x0

Fig. 5.8 Graphical representation of the Dyson equation for the two-particle Green’s function (5.80). The crossed box represents the proper two-particle self-energy

Similarly, the Dyson equation for two-particle Green’s function becomes G.x; x

0

I x0 ; x00 /

D G0 .x; x

0

I x0 ; x00 /

ZZZZ C

d4 x1 d4 x2 d4 x10 d4 x20

G0 .x; x 0 I x2 ; x20 / .i/˙ .x2 ; x20 I x1 ; x10 / G.x1 ; x10 I x0 ; x00 /:

(5.80)

This equation is illustrated in Fig. 5.8, where the crossed box represents the proper two-particle self-energy.

110

5 Green’s Functions Table 5.1 Electron affinity of Ca atom (in meV) 4p1=2 4p3=2 Reference Theory 19 13 Salomonson [11] Theory 22 18 Avgoustoglou [3] Theory 49 18 Dzuba [3] Expt’l Expt’l Expt’l

24.55 18.4 17.5

19:73

Petrunin [10] Walter [13] Nadeau [9]

5.3.4 Numerical Illustration Here, we shall illustrate the application of the Green’s-function technique for manybody calculation by the electron affinity of the calcium atom (Table 5.1). The negative calcium ion is a very delicate system, with a very feeble binding energy, and it has been quite difficult to determine this quantity experimentally as well as theoretically. It is only recently that it has been possible to obtain reasonable agreement. The calculation of Salomonson et al. is performed by means of the Green’sfunction method, that of Dzuba et al. by many-body perturbation theory and that of Avgoustoglou by all-order pair-correlation method.

5.4 Field-Theoretical Green’s Function: Open-Shell Case In this section, we indicate how the Green’s-function concept could be extended to the open-shell case, when the model states are separated from the vacuum state. It is recommended that Chap. 6 is first studied, where the treatment is more akin to the normal situation in MBPT, discussed in Sect. 2.3. We leave out most details here and refer to the treatment of the covariant evolution operator and the Green’s operator, which is quite equivalent. In this section, we +in particular look into the special approach due to Shabaev [12].

5.4.1 Definition of the Open-Shell Green’s Function In the general open-shell case, singularities of the Green’s function can appear also for connected diagrams, as in the covariant-evolution operator (see below). If we consider a sequence of ladder diagrams of single-photon exchange, V , as discussed in the next chapter (Fig. 6.3), considering only particle states (no-pair), the Feynman amplitude for the Green’s function is the same as for the covariant evolution operator (6.20) with no model-space states, M D 1 C Q .E0 / V .E0 / C Q .E0 / V .E0 / Q .E0 / V .E0 / C ;

(5.81)

5.4 Field-Theoretical Green’s Function: Open-Shell Case

111

where Q .E0 / D

Q jrsi hrsj D E0 H0 C i E0 "r "s C i

is the reduced resolvent (2.65) and E0 is the energy parameter (of the Fourier transform) of the Green’s function. The GF becomes singular, when there is an intermediate statejrsi of energy E0 . Including the residuals after removing the singularities (model-space contributions) leads as shown below (6.117) to a shift of the energy parameter, E0 ! E D E0 C E, M D 1 C Q .E/ V .E/ C Q .E/ V .E/ Q .E/ V .E/ C :

(5.82)

This is a Brillouin–Wigner perturbation expansion, and it can be summed to MD

jni hnj 1 D E H C i E En C i

(5.83)

with H D H0 CV .E/ andjni represents the exact eigenstates of the system with the energy En . This agrees with the Fourier transform of the GF derived above (5.47), demonstrating that the transform has poles at the exact energies. Consequently, this holds also in the open-shell case. The Green’s-function technique yields information only about the energy of the system. This is in contrast to the Green’s-operator formalism, to be treated in the next chapter, which can give information also about the wave function or state vector of the system under study.

5.4.2 Two-Times Green’s Function of Shabaev The use of the Green’s-function technique for atomic calculations has been further developed by Shabaev et al. [12] under the name of the “Two-times Green’s function” (which is equivalent to the equal-time approximation, discussed above). This technique is also applicable to degenerate and quasi-degenerate energy states, and we shall outline its principles here. We return to the extended-model concept, discussed in Sect. 2.3. Given are a number of eigenstates (target states) of the many-body Hamiltonian Hj ˛ i D E ˛j ˛ i

.˛ D 1 d /:

(5.84)

The corresponding model states are in intermediate normalization the projections on the model space j0˛ i D P j ˛ i

.˛ D 1 d /:

(5.85)

112


The model states are generally nonorthogonal, and following Shabaev we ˚ˇ ˇ ˛

0 , defined by introduce a “dual set” ˇe ˇ E ˇ ED ˇ ˇ eˇ ˇ ˇ e˛ ˇ ˇ 0 h0˛ jD ˇ0 0 ˇ D ı˛;ˇ :

(5.86)

Then the standard projection operator becomes P D

X ˇˇ ˇ E D ˇ ˇˇ X ˇˇ ˇ E D ˇ ˇˇ e ˇD e ˇ ˇ ˇ0 0 0 0 ˇ 2D

(5.87)

ˇ 2D

with the summation performed over the model space D. We also define an alternative projection operator as PD

X ˇˇ ˇ E D ˇ ˇˇ ˇ0 0 ˇ

P 1 D

ˇ 2D

Then

X ˇˇ ˇ E D ˇ ˇˇ e e ˇ: ˇ 0 0

(5.88)

ˇ

ˇ E ˇ E ˇe˛ ˇ ˛ P ˇ 0 D ˇ0

and

ˇ E ˇ E ˇ ˇe ˛ P 1 ˇ0˛ D ˇ 0 :

(5.89)

The Fourier transform of the retarded Green’s function is generally (5.33) GC .EI x; x 0 / D i

X hxjn i hn jx 0 i n

E En C i

;

(5.90)

where we let x; x 0 represent the space coordinates of all outgoing/incoming particles. It then follows that I dE GC .EI x; x 0 / D 2 hxjn i hn jx 0 i (5.91) n

and

I n

E dE GC .EI x; x 0 / D 2 hxjn i En hn jx 0 i;

(5.92)

where n is a closed contour, encircled in the positive direction and containing the single target energy En and no other pole. (This holds if all poles are distinct. In the case of degeneracy, we can assume that an artificial interaction is introduced that lifts the degeneracy, an interaction that finally is adiabatically switched off.) This yields the relation [12, Eq. (44)] H En D Hn

E dE GC .EI x; x 0 /

n

dE GC .EI x; x 0 /

:

(5.93)


113

Following Shabaev, we also introduce a “projected” Green’s function by D ˇ ED ˇ E ˇ ˇ ˇˇ X x ˇ0 0 ˇx 0 ; (5.94) gC .EI x; x 0 / D i E E ˇ C i ˇ 2D

which is the coordinate representation (see Appendix C.3) of the corresponding operator ˇ ˛˝ ˇ ˇ ˛˝ ˇ X P ˇ ˇ ˇ ˇP X ˇ0ˇ 0ˇ ˇ gO C .E/ D i Di (5.95) E E ˇ C i E E ˇ C i ˇ 2D

ˇ 2D

operating only within the model space. The effective Hamiltonian (2.54) is defined by Heffj0˛ i D E ˛j0˛ i and we can then express this operator as Heff D

X ˇˇ ˇ E D ˇ ˇˇ e ˇ D Heff P 1 ; ˇ0 E ˇ 0

(5.96)

ˇ 2D

where Heff D

X

ˇ E D ˇ ˇ ˇ ˇˇ D ˇ0 E ˇ 0 ˇ:

(5.97)

ˇ 2D

From the definition (5.94), it follows that I 1 Heff D E dE g.E/; 2

(5.98)

where the integration contour contains the energies all target states. As before, we assume that the poles are distinct. Expanding the effective Hamiltonian (5.96) order-by-order leads to .0/ .1/ .0/ .1/ C Heff Heff P C : Heff D Heff

The first-order operator Heff becomes .1/

Heff D

1 2

where .0/ gC .EI x; x 0 / D i

I

(5.99)

.0/

E dE gC .E/; D ˇ ED ˇ E ˇ ˇ ˇˇ X x ˇ0 0 ˇx 0 ˇ 2D

E E ˇ C i

(5.100)

:

(5.101)

114


The effective Hamiltonian above is nonhermitian, as in the MBPT treatment in Sect. 2.3. It can also be given a hermitian form [12], but we shall maintain the nonhermitian form here, since it makes the formalism simpler and the analogy with the later treatments more transparent.

5.4.3 Single-Photon Exchange We shall now apply the two-times Greens function above to the case of singlephoton exchange between the electrons, discussed above (Fig. 5.2). We shall evaluate the contribution to the effective Hamiltonian in the general quasi-degenerate case. In the equal-time approximation, the (first-order) Green’s function (5.25) is given by 0 0 it .!3 C!4 / it0 .!1 C!2 / G .1/ .x; x 0 ; x0 ; x00 / D M.1/ e (5.102) sp .x; x I x 0 ; x 0 / e

and the first-order Feynman amplitude is given by (5.23) 0 0 M.1/ sp .x; x I x 0 ; x 0 /

ZZ D i

d!3 d!4 2 2

ZZ

d!1 d!2 SF .!3 I x; x 1 / 2 2

SF .!4 I x 0 ; x 2 / I.!1 !3 I x 2 ; x 1 / SF .!1 I x 1 ; x 0 / SF .!2 I x 2 ; x 00 / 22 .!1 C !2 !3 !4 /; (5.103) after integrations over z. The Fourier transform of the Green’s function with respect to t and t0 is G

.1/

ZZ

0

.E ; E/ D

dt dt0 iE 0 t iE t0 .1/ e G .x; x 0 ; x0 ; x00 / e 2 2

0 0 D .E 0 !3 !4 / .E!1 !2 / M.1/ sp .x; x I x 0 ; x 0 /

(5.104)

or G

.1/

0

ZZ

d!3 d!1 SF .!3 I x; x 1 / SF .E 0 !3 I x 0 ; x 2 / 2 2 I.!1 !3 I x 2 ; x 1 / SF .!1 I x 1 ; x 0 / SF .E !1 I x 2 ; x 00 / 22 .E 0 E/; (5.105)

.E ; E/ D i

after integrations over !2 ; !4 . With the expression for the electron propagator (4.12), the matrix element of the Green’s function becomes


115

ˇZ Z ˇ ˛ ˇ ˝ ˇ .1/ 0 1 1 d!3 d!1 rs ˇG .E ; E/ˇtu D rs ˇˇ 2 2 !3 "r C i u E 0 !3 "s C i s ˇ ˇ 1 1 ˇ tu I.!1 !3 / !1 "t C i t E !1 "u C i u ˇ 22 .E 0 E/: (5.106)

With jrsi and jtui in the model space, this is the same as the matrix element of the projected Green’s function (5.94), considering only poles corresponding to the relevant target states. We define the single-energy Fourier transform by Z G.E/ D

dE 0 G.E 0 ; E/; 2

(5.107)

which yields ˇZ Z ˇ ˛ ˇ ˝ ˇ .1/ d!3 d!1 ˇ ˇ I.!1 !3 / rs G .E/ tu D i rs ˇˇ 2 2

1 1 1 C E "r "s !3 "r C i r E !3 "s C i s

ˇ ˇ 1 1 1 ˇ tu : C E "t "u !1 "t C i t E !1 "u C i u ˇ (5.108) We assume here that the initial and final states lie in the model space with all single-particle states involved being particle states. The relevant poles are here E D "t C "u D Ein

and E D "r C "s D Eout :

The contribution of the first pole is ˇZ Z ˇ i d!3 d!1 rs ˇˇ I.!1 !3 / 2i 2 2

Ein 1 1 C Ein Eout !3 "r C i Ein !3 "s C i ˇ

ˇ 1 1 ˇ tu : C !1 "t C i Ein !1 "u C i ˇ

(5.109)

The last bracket yields 2i .!1 "t /, and integration over !1 yields ˇZ ˇ

ˇ ˇ Ein 1 1 ˇ tu : i rs ˇˇ !3 I."t !3 / C Ein Eout !3 "r C i Ein !3 "s C i ˇ (5.110)

116


Similarly, the other pole yields ˇZ ˇ d!1 I.!1 "r / i rs ˇˇ 2 ˇ

ˇ Eout 1 1 ˇ tu : (5.111) C Ein Eout !1 "t C i Eout !1 "u C i ˇ The matrix element of P .1/ is similar with Ein and Eout in the numerator re.0/ moved. The matrix element of Heff P .1/ is obtained by multiplying by Eout , and the first-order contribution then becomes ˇZ ˇ E D ˇ ˇ d!3 ˇ .1/ ˇ rs ˇHeff I."t !3 / ˇ tu D i rs ˇˇ 2 ˇ

ˇ 1 1 ˇ tu : (5.112) C !3 "r C i Ein !3 "s C i ˇ The photon interaction is in the Feynman gauge given by (4.46) Z I.qI x 1 ; x 2 / D

2c 2 d f F .I x 1 ; x 2 / q 2 c 2 2 C i

with f F given by (4.55). This gives Z 2c 2 d f F .I x 1 ; x 2 / I."t !3 / D ."t !3 /2 c 2 2 C i

(5.113)

(5.114)

with the poles at !3 D "t ˙ .i/. Integrating the relation (5.112) over !3 then yields .1/ hrsjHeff jtui

ˇZ

ˇ F ˇ D rs ˇ cd f

ˇ ˇ 1 1 ˇ tu : C "t "r . i/ "u "s . i/ ˇ (5.115)

This agrees with the result obtained with the covariant evolution operator (CEO) method in the next chapter (6.16). The CEO result is more general, since it is valid also when the initial and/or final states lie in the complementary Q space, in which case the result contributes to the wave function or wave operator. In contrast to the S-matrix formulation, the Green’s-function method is applicable also when the initial and final states have different energies, which makes it possible to evaluate the effective Hamiltonian in the case of an extended model space and to handle the quasi-degenerate case. The two-times Green’s function has in recent years been successfully applied to numerous highly charged ionic systems by Shabaev, Artemyev et al. of the St. Petersburg group for calculating two-photon radiative effects, fine structure separations, and g-factors of hydrogenic systems [1, 2, 14, 15]. Some numerical results are given in Chap. 7.

References

117

References 1. Artemyev, A.N., Beier, T., Plunien, G., Shabaev, V.M., Soff, G., Yerokhin, V.A.: Vacuumpolarization screening corrections to the energy levels of heliumlike ions. Phys. Rev. A 62, 022,116.1–8 (2000) 2. Artemyev, A.N., Shabaev, V.M., Yerokhin, V.A., Plunien, G., Soff, G.: QED calculations of the n=1 and n=2 energy levels in He-like ions. Phys. Rev. A 71, 062,104 (2005) 3. Avgoustoglou, E.N., Beck, D.R.: All-order relativistic many-body calculations for the electron affinities of Ca , Sr , Ba , and Yb negative ions. Phys. Rev. A 55, 4143–49 (1997) 4. Bjorken, J.D., Drell, S.D.: Relativistic Quantum Mechanics. Mc-Graw-Hill Pbl. Co, N.Y. (1964) 5. Fetter, A.L., Walecka, J.D.: The Quantum Mechanics of Many-Body Systems. McGraw-Hill, N.Y. (1971) 6. Feynman, R.P.: The Theory of Positrons. Phys. Rev. 76, 749–59 (1949) 7. Itzykson, C., Zuber, J.B.: Quantum Field Theory. McGraw-Hill (1980) 8. Mahan, G.D.: Many-particle Physics, second edition. Springer Verlag, Heidelberg (1990) 9. Nadeau, M.J., Zhao, X.L., Garvin, M.A., Litherland, A.E.: Ca negative-ion binding energy. Phys. Rev. A 46, R3588–90 (1992) 10. Petrunin, V.V., Andersen, H.H., Balling, P., Andersen, T.: Structural Properties of the Negative Calcium Ion: Binding Energies and Fine-Structure Splitting. Phys. Rev. Lett. 76, 744 (1996) 11. Salomonson, S., Warston, H., Lindgren, I.: Many-Body Calculations of the Electron Affinity for Ca and Sr. Phys. Rev. Lett. 76, 3092–95 (1996) 12. Shabaev, V.M.: Two-times Green’s function method in quantum electrodynamics of high-Z few-electron atoms. Physics Reports 356, 119–228 (2002) 13. Walter, C., Peterson, J.: Shape resonance in Ca photodetachment and the electron affinity of Ca(1 S). Phys. Rev. Lett. 68, 2281–84 (1992) 14. Yerokhin, V.A., Artemyev, A.N., Beier, T., Plunien, G., Shabaev, V.M., Soff, G.: Two-electron self-energy corrections to the 2p1=2 2s transition energy in Li-like ions. Phys. Rev. A 60, 3522–40 (1999) 15. Yerokhin, V.A., Artemyev, A.N., Shabaev, V.M., Sysak, M.M., Zherebtsov, O.M., Soff, G.: Evaluation of the two-photon exchange graphs for the 2p1=2 2s transition in Li-like Ions. Phys. Rev. A 64, 032,109.1–15 (2001)

Chapter 6

Covariant Evolution Operator and Green’s Operator

The third method we shall consider for numerical QED calculation on bound states is the covariant-evolution-operator (CEO) method, developed during the last decade by the Gothenburg group [9]. This procedure is based upon the nonrelativistic timeevolution operator, discussed in Chap. 3, but it is made covariant in order to be applicable in relativistic calculations. Later, we shall demonstrate that this procedure forms a convenient basis for a covariant relativistic many-body perturbation procedure, including QED as well as correlational effects, which for two-electron systems is fully compatible with the Bethe–Salpeter equation. This question will be the main topic of the rest of the book.

6.1 Definition of the Covariant Evolution Operator In the standard time-evolution operator (3.6), U.t; t0 /, time is assumed to evolve only forwards in the positive direction, which implies that t t0 . Internally, time may run also backwards in the negative direction, which in the Feynman/Stückelberg interpretation [6, 15] represents the propagation of hole or antiparticle states with negative energy. However, all internal times (ti ) are limited to the interval ti 2 Œt; t0 . In the S-matrix (4.2), the initial and final times are t0 D 1 and t D C1, respectively, which implies that the internal integrations do run over all times, making the concept Lorentz covariant.1 In order to make the time-evolution operator covariant also for finite times, it has to be modified. This leads to what is referred to as the covariant evolution operator (CEO), introduced by Lindgren, Salomonson, and coworkers in early 2000s [7–11]. The CEO is, as well as the S-matrix and the Green’s function, field-theoretical concepts, and the perturbative expansions of these objects are quite similar. The integrations are performed over all times, and therefore, these objects are normally represented by Feynman diagrams instead of time-ordered Goldstone diagrams, discussed earlier (Sect. 2.4). 1

See footnote in the Introduction.


119

120

6 Covariant Evolution Operator and Green’s Operator

The evolution operator contains generally (quasi)singularities, when it is unlinked or when an intermediate state lies in the model space. Later in this chapter, we see how these singularities can be removed for the CEO, leading to what we refer to as the Green’s operator, since it is quite analogous to the Green’s function, which is also free of singularities. As mentioned earlier, the covariant perturbation expansion we shall formulate here leads for two-particle systems ultimately to the full Bethe–Salpeter (BS) equation [10]. In principle, the BS equation has separate time variables for the individual particles, which makes it manifestly covariant. This is also the case for the CEO as well as for the Green’s function. In most applications, however, times are equalized, so that the objects depend only on a single time, which is known as the equal-time approximation. This makes the procedure in line with the standard quantum-mechanical picture, where the wave function has a single time variable, .t; x1 ; x 2 /, but the covariance is then partly lost. Here, we shall mainly work with this approximation in order to be able to combine the procedure with the standard many-body perturbation theory. As a first illustration, we consider the single-photon exchange with the standard evolution operator (Fig. 6.1, left), the Green’s function (middle), and the CEO (right). In the standard evolution operator only particle states (positive-energy states) are involved in the lines in and out. Therefore, this operator is NOT Lorentz covariant. In the Green’s function, there are electron propagators on the free lines, involving particle as well as hole states (positive- and negative-energy states), and the internal times can flow in both directions between 1 and C1, which makes the concept covariant. In the CEO electron propagators are inserted on the free lines of the standard evolution operator with integration over the space coordinates, making it covariant. This implies that we attach a density operator [3] to the free lines .x/ O D O .x/ O .x/

(6.1)

with integration over the space coordinates. We can also see the CEO as the Green’s function, with electron-field operators attached to the free ends.

t

- r Particles 1

-

t t0

x

t OC

z

-

OC

OC

s 2

OC

u

- r Part. Holes

t0

1 t

O˙

z

-

O˙

O˙

s 2

O˙

u

r 6O ˙

- r Part. Holes

x0

1 t

O˙

z

-

O˙

t 6O ˙

O

˙

6s

O˙

s

x0

2 O˙

u

O ˙ 6u

x00

Fig. 6.1 Comparison between the standard evolution operator, the Green’s function, and the covariant evolution operator for single-photon exchange in the equal-time approximation

6.1 Definition of the Covariant Evolution Operator

121

We generally define the Covariant Evolution Operator (CEO) in the single-

particle case by the one-body operator2 ZZ 1 UCov .t; t0 / D

ˇ ˛ ˝ ˇ d3 x d3 x 0 O .x/ 0H ˇT Œ O H .x/ O H .x0 /ˇ0H O .x0 /:

(6.2)

We use here the same vacuum expectation in the Heisenberg representation as in the definition of the Green’s function (5.8) with two additional electron-field operators, O .x/ and O .x0 /, with space integrations over x; x 0 . In contrast to the Green’s function, we shall assume here that the number of photons does not need to be conserved. When this number is conserved, the vacuum expectation is a number and represents the corresponding Green’s function. The space integration makes the electron-field operators attached to this function, as illustrated in the figure below (c.f. Fig. 5.1, left). O x SF

x0

SF O

In analogy with the expression (5.18) for the Green’s function, we can also ex-

press the covariant evolution operator as ZZ 1 UCov .t; t0 / D

d3 x d3 x 0 .x/ O U 1 .1; 1/ .x O 0 /;

(6.3)

where the density operators are connected to the standard (one-body) evolution operator or S-matrix. In expanding the S-matrix [see (5.16)], we obtain 1 .t; t0 / D UCov

Z ZZ 1 i n Z X 1 d3 x d 3 x 0 d4 x1 d4 xn nŠ c nD0 h i T .x/ O H.x1 / H.xn / .x O 0 / e.jt1 jCjt2 j / ; 1

(6.4)

where the operators are connected to form a one-body operator.

2

An “n-body operator” is an operator with n pairs of creation/absorption operators (for particles), while an “m-particle” function or operator is an object of m particles outside our vacuum. In principle, n can take any value n m, although we shall normally assume that n D m.

122


Similarly, the two-particle CEO becomes – in analogy with the corresponding

Green’s function (5.21) and Fig. 5.1 (right) ZZZZ 1 i n X 1 D d3 x d3 x 0 d3 x 0 d3 x 00 nŠ c nD0 Z Z h d4 x1 d4 xn T .x/ O .x O 0 / H.x1 / H.xn /

2 UCov .t; t 0 I t0 ; t00 /

.x O 00 /.x O 0/

x

O

2

e.jt1 jCjt2 j / :

(6.5)

O x0 SF

SF

x0

i

SF

SF x0 O 0

O

6.2 Single-Photon Exchange in the Covariant-Evolution-Operator Formalism We shall now consider the exchange of a single photon between the electrons in the covariant-evolution-operator formalism. We consider here a general covariant gauge (see Sect. 4.3), such as the Feynman gauge, and we shall later consider the noncovariant Coulomb gauge. We assume here that the initial state is unperturbed and return to the more general situation in Chap. 8. The CEO for the exchange of a single photon (Fig. 6.1, rightmost) is in the general case given by Usp .t; t

0

I t0 ; t00 /

ZZ D

3 0

d xd x 3

(

ZZ d

3

x 0 d3 x 00

O

.x/

O

0

.x /

1 2

ZZ d4 x1 d4 x2

iSF .x; x1 / iSF .x 0 ; x2 / .i/ e 2 DF .x2 ; x1 / iSF .x1 ; x0 / ) iSF .x2 ; x00 / e.jt1 jCjt2 j/ O .x00 / O .x0 /

(6.6)

6.2 Single-Photon Exchange in the Covariant-Evolution-Operator Formalism

123

[with DF defined in (4.19)] in analogy with the corresponding S-matrix and Green’s-function expressions. The expression in the curly brackets is the corresponding Green’s function (5.22) (the denominator does not contribute in first order). The CEO contains additional electron creation/annihilation operators and integration over the space coordinates at the initial and final times. This makes the CEO into an operator, while the Green’s function is a function. When the initial state is unperturbed, it implies with the adiabatic damping that the initial time (t0 ; t00 : : :) is 1. From the definition of the electron propagator (4.8), it can be shown that, as t0 ! 1, Z (6.7) d3 x 0 iSF .x; x0 / O .x0 / ) O .x/; when the incoming state is a particle state. Therefore, we can leave out the propagators on the incoming lines, as illustrated in the first diagram of Fig. 6.2, corresponding to the expression (

ZZ

0

Usp .t; t I 1/ D

3 0

d xd x 3

O

.x/

0

O

1 .x / 2 0

ZZ d4 x1 d4 x2 i SF .x; x1 / )

i SF .x ; x2 / .i/e DF .x2 ; x1 / e 2

.jt1 jCjt2 j/

O .x2 / O .x1 /: (6.8)

Identification with the expression for the second quantization (Appendix B) leads to the matrix element ZZ ˝ ˝ ˇ 0 i.t "r Ct 0 "s / dt1 dt2 rs ˇx; x 0 i x; x 0 jiSF .x; x1 / hrsjUsp .t; t I 1/jabi D e ˇ ˛ iSF .x 0 ; x2 / .i/e 2 DF .x2 ; x1 /jx 1 ; x 2 i hx 1 ; x 2 jˇab ei.t1 "a Ct2 "b / e.jt1 jCjt2 j/ ; where we for clarity have indicated the integration variables (see Appendix C.3).

x

r 6O r

!1

z

!2

-

1 a

O

O

s

x0 x

r 6

b

6s s

r

x0 x

r 6

a

6s

r

-

2 O

E0

6s

s

b

a

E0

Fig. 6.2 The evolution-operator diagram for single-photon exchange

b E0

x0

124

6 Covariant Evolution Operator and Green’s Operator 0

The external-time dependence is here eit .!1 "r / eit .!2 "r / , which in the equal-time approximation (t D t 0 ) becomes eit .!1 C!2 "r "s / . Since in the limit ! 0 !1 C !2 D "a C "b D E0 is the initial energy and "r C "s is the final energy, we have in this limit hrsjUsp .t; 1/jabi D eit .E0 "r "s / hrsjMspjabi

(6.9)

Usp .t; 1/jabi D eit .E0 H0 / jrsi hrsjMspjabi ;

(6.10)

or

where Msp represents the Feynman amplitude. This is defined as the operator without the external time dependence, in analogy with the Green’s function (5.23) (see also Appendix H.2). This yields ˇZ Z Z ˇ ˝ ˛ d!1 d!2 dz rsjMsp jab D rs ˇˇ iSF .!1 I x; x 1 / iSF .!2 I x 0 ; x 2 / 2 2 2 .i/I.zI x 2 ; x 1 /2 ."a z !1 / ˇ ˇ (6.11) 2 ."b C z !2 /ˇˇab and after integration over !1 ; !2 in the limit ! 0 D ˇ Z dz ˇ iSF ."a zI x; x 1 / iSF ."b C zI x 0 ; x 2 / hrsjMspjabi D rs ˇ 2

ˇ E ˇ .i/I.zI x 2 ; x 1 / 22 ."b C "b !1 !2 / ˇab :

(6.12)

Inserting the expressions for the propagator (4.10) and the interaction (4.46) then yields ˇ Z ˇ 1 dz 1 hrsjMspjabi D rs ˇˇ i 2 "a z "r C ir "b C z "s C is ˇ Z 2c 2 d f . / ˇˇ ab : z2 c 2 2 C i ˇ

(6.13)

With the identity (4.75), this can be expressed hrsjMspjabi D

1 hrsjVspjabi E0 "r "s

or Msp .x; x 0 /jabi D

1 Vspjabi ; E0 H0

(6.14)

(6.15)

6.2 Single-Photon Exchange in the Covariant-Evolution-Operator Formalism

125

where Vsp is the potential for single-photon exchange (4.77), ˇZ 1 ˇ c d f . / hrsjVspjabi D rs ˇˇ 0 ˇ ˇ 1 1 ˇab : C "a "r .c i /r "b "s .c i /s ˇ (6.16) The evolution operator (6.10) then becomes ˇ ˛ eit .E0 H0 / Usp .t; 1/ˇab D Vspjabi : E0 H0

(6.17)

The results above hold in any covariant gauge, like the Feynman gauge. They do hold also for the transverse part in the Coulomb gauge by using the transverse part of the f function (4.60). The result (6.16) is identical to the Green’s-function result (5.115), when the final state,jrsi, lies in the model space. In the CEO case, the final state can also lie in the complementary Q space, in which case the evolution operator contributes to the wave function/operator. The CEO result can be represented by means of two time-ordered Feynman diagrams, as shown in Fig. 6.2. We then see that the denominators are given essentially by the Goldstone rules of standard many-body perturbation theory [5, Sect. 12.4], i.e., the unperturbed energy minus the energies of the orbital lines cut by a horizontal line, in the present case including also k for cutting the photon line.3 When the initial and final states have the same energy, the potential (6.16) above becomes ˇ ˇZ 1 ˇ 2 d f . / ˇˇ ab ; (6.18) hcd jVspjabi D cd ˇˇ q 2 2 C i ˇ 0 where cq D "a "c D "d "b , which is the energy-conservative S-matrix result (4.46) and (4.52). We have seen here that the covariant evolution operator for single-photon exchange has the time dependence eit .E0 H0 / , which differs from that of the nonrelativistic evolution operator (3.11). We shall return to this question at the end of this chapter.

3

It should be observed that a Goldstone diagram is generally distinct from a “time-ordered Feynman diagram,” as is further analyzed in Appendix I.

126


6.2.1 Single-Photon Ladder We can now construct the covariant evolution operator for some ladder-type interactions provided no hole states are appearing in the intermediate states (see Chap. 8). The diagram in Fig. 6.3 represents two reducible single-photon interactions with an intermediate time (t 0 ) that separates the interactions. The Feynman amplitude is then obtained by combining two single-photon interactions (6.16) with corresponding resolvents, MPE D .E/ Vsp .E/ .E/ Vsp .E/PE : (6.19) where (see 8.11) Z ˇ ˛ D ˇˇ ˝ ˇ rs ˇVsp .E/ˇtu D rs ˇ

h

1

c d f . / 0

iˇ E 1 1 ˇ C ˇtu E "t "s c E "r "u c

PE is the projection operator of the part of the model space with energy E, and .E/ is the resolvent (2.64). This procedure can be repeated to a general single-photon ladder MLadd PE D .E/ Vsp .E/ .E/ Vsp .E/ .E/ Vsp .E/PE :

(6.20)

The corresponding part of the evolution operator is according to (6.10) U0 .t; 1/Ladd PE D eit .EH0 / MLadd PE ;

(6.21)

where subscript “0” is used to indicate that there are no intermediate model-space states (see further below). This evolution operator can be singular due to intermediate and/or final model-space states, which can be eliminated by means of counterterms, leading to “folds” (model-space contributions, MSC), as we shall demonstrate below. It should be observed that In the equal-time approximation the interactions and the resolvents as well as

the time factor of the ladder without folds all depend on the energy of the initial, unperturbed state. The folds will affect the time dependence, as will be discussed in Sect. 6.9. In Part III, we shall treat the ladder in the presence of virtual pairs and higher-order interactions and see how the procedure can be fitted into a many-body procedure.

t

t0 Fig. 6.3 Feynman diagram representing second-order ladder diagram (6.19)

6

6

66

6

6

6

6

-

E

6

6

6.3 Multiphoton Exchange

127

6.3 Multiphoton Exchange 6.3.1 General We shall now briefly consider the general case of multiphoton exchange. We can describe this by means of a general many-body potential, which we can separate into one-, two-, . . . body parts, V D V1 C V2 C V3 C

(6.22)

and which contains all irreducible interactions.4 By iterating such a potential, all reducible interactions will be generated. In Figs. 6.4 and 6.6, we illustrate the one- and two-body parts of this potential, including radiative effects – vacuum polarization, self-energy, vertex correction (see Sect. 2.6) – which, of course, have to be properly renormalized (see Chap. 12). The one-body potential contains an effective-potential interaction (Fig. 6.5) in analogy to that in ordinary MBPT (2.73). In the effective potential here, however, the internal lines can be hole lines as well as particle lines. This implies that the second diagram on the r.h.s. in Fig. 6.5 contains the direct Hartree–Fock potential as well as the radiative effect of vacuum polarization and the last diagram the exchange

V1 x

=

+

+

+

Fig. 6.4 Graphical representation of the one-body part of the effective potential (6.22), containing the one-body potential in Fig. 6.5 as well as irreducible one-body potential diagrams, including radiative effects

=

+

?+

Fig. 6.5 Graphical representation of the “extended” effective potential interaction. This is analogous to the effective potential in Fig. 2.3, but the internal lines represent here all orbitals (particles as well as holes). This implies that the last two diagrams include the (renormalized) vacuum polarization and self-energy

4

Concerning the definition of the concepts “reducible” and “irreducible,” see Sect. 2.6.

128


V2

=

Fig. 6.6 The two-body part of the effective potential (6.22) contains all irreducible two-body potential diagrams

part of the HF potential as well as the and electron self-energy (both radiative effects properly renormalized). All heavy lines here represent orbitals in the external (nuclear) potential, which implies that the vacuum polarization contains the Uehling potential [16] (see Sect. 4.6) as well as the Wickmann–Kroll [17] correction, discussed earlier in Sect. 4.6.

6.3.2 Irreducible Two-Photon Exchange We consider next the general two-photon exchange, illustrated in Fig. 6.7, still assuming the equal-time approximation and unperturbed initial state.

6.3.2.1 Uncrossing Photons Generalizing the result for single-photon exchange (6.6), we find that the kernel of the first (ladder)diagram becomes iSF .x; x3 / iSF .x 0 ; x4 / .i/e 2 DF .x4 ; x3 / iSF .x3 ; x1 / iSF .x4 ; x2 / .i/e 2 DF .x2 ; x1 /:

(6.23)

This leads to the Feynman amplitude in analogy with (6.11) Msp .x; x 0 I x 0 ; x 00 / D

ZZZZ

d!1 d!2 d!3 d!4 2 2 2 2

ZZ

dz dz0 iSF .!3 I x; x 3 / 2 2

iSF .!4 I x 0 ; x 4 / .i/I.z0 I x 4 ; x 3 / iSF .!1 I x 3 ; x 1 / iSF .!2 I x 4 ; x 2 / .i/I.zI x 2 ; x 1 / 2 ."a !1 z/

6.3 Multiphoton Exchange

x

129

r 6 r 3 t 1 a

6s

!3

0

z -

!1

z

-

!4

s

!2

4 u

E

x0

x

r 6

6s

r !3 !4 4 z0 t !1 !2 z6 1 a

2 b

E

x0

s 2 u 3 b

Fig. 6.7 Covariant-evolution-operator diagrams for two-photon ladder and “cross”

2 ."b !2 C z/ 2 .!1 z0 !3 / 2 .!2 C z0 !4 /:

(6.24)

Integration over !1 ; !2 leads to Msp .x; x 0 / D

ZZ

d!3 d!4 2 2

ZZ

dz dz0 iSF .!3 I x; x 3 / iSF .!4 I x 0 ; x 4 / 2 2

.i/I.z0 I x 4 ; x 3 / iSF ."a zI x 3 ; x 1 / iSF ."b C zI x 4 ; x 2 / .i/I.zI x 2 ; x 1 / 22 ."a z z0 !3 / 22 ."b C z C z0 !4 /

(6.25)

and over !3 ; !4 Msp .x; x 0 / D

ZZ

dz dz0 iSF ."a z z0 I x; x 3 / iSF ."b C z C z0 I x 0 ; x 4 / 2 2

.i/I.z0 I x 4 ; x 3 / iSF ."a zI x 3 ; x 1 / iSF ."b C zI x 4 ; x 2 / .i/I.zI x 2 ; x 1 /: Integration over z0 leads to the denominators 1 1 1 C E "r "s "a "r z .c 0 i /r "b "s C z .c 0 i /s and the remaining part of the integrand is 1 1 1 1 C : 2 2 E "t "u "a "t z C it "b C z C iu z c 2 C i

(6.26)

130


6.3.2.2 Crossing Photons For the crossed-photon exchange in Fig. 6.7 (right), the corresponding result is Msp .x; x

0

I x 0 ; x 00 /

ZZZZ D

d!1 d!2 d!3 d!4 2 2 2 2

ZZ

dz dz0 iSF .!3 I x; x 4 / 2 2

iSF .!4 I x 0 ; x 2 / .i/I.z0 I x 4 ; x 3 / iSF .!1 I x 4 ; x 1 / iSF .!2 I x 2 ; x 3 / .i/I.zI x 2 ; x 1 / 2 ."a !1 z/ 2 ."b !2 z0 / 2 .!1 z0 !3 / 2 .!2 C z !4 /:

(6.27)

Integration over the omegas yields Msp .x; x 0 / D

ZZ

dz dz0 iSF ."a C z0 zI x; x 4 / iSF ."b C z z0 I x 0 ; x 2 / 2 2

.i/I.z0 I x 4 ; x 3 / iSF ."a zI x 4 ; x 1 / iSF ."b zI x 2 ; x 3 / .i/I.zI x 2 ; x 1 /:

(6.28)

Integration over z0 leads to the denominators 1 E "r " s

1 1 C "a "r z .c 0 i /r "b "s C z .c 0 i /s

and the remaining part of the integrand is 1 1 1 : 2 2 "a "t z C it "b "u z C iu z c 2 C i

To evaluate the integrals above is quite complicated, but they are considered in ˚ en [2]. The two-photon efdetail in [9, Appendix B] and in the thesis of Björn As´ fects have been evaluated for helium-like ions, and some results are shown in the following chapter.

6.3.3 Potential with Radiative Parts Two-photon potentials with self-energy and vacuum-polarization insertions can also be evaluated in the covariant-evolution-operator formalism, as discussed in [9]. We shall not consider this any further here, but return to these effects in connection with the MBPT–QED procedure in Chap. 8.

6.4 Relativistic Form of the Gell-Mann–Low Theorem

131

6.4 Relativistic Form of the Gell-Mann–Low Theorem We have in Chap. 3 considered the nonrelativistic form of the Gell-Mann–Low theorem, and we shall now extend this to the relativistic formalism. This theorem plays a fundamental role in the formalism we shall develop here. We shall start with a conjecture that the time evolution of the relativistic state vector is governed by

the CEO in the equal-time approximation (in the interaction picture), in analogy with the situation in the nonrelativistic case (3.6) (c.f. [4, Sect. 6.4]), ˇ ˛ ˇ ˛ ˛ ˇ .t/ D UCov .t; t0 /ˇ ˛ .t0 / : Rel Rel

(6.29)

We shall later demonstrate that this conjecture is consistent with the standard quantum-mechanical picture (6.120) [see also (9.13)]. It should be noted that the evolution operator does not generally preserve the (intermediate) normalization. It can now be shown as in the nonrelativistic case in Sect. 3.3 that the conjecture above leads to a relativistic form of the Gell-Mann–Low theorem for a general quasi-dege-

nerate model space ˇ ˛ ˛ ˇ ˛ ˇ .0/ D ˇ ˛ D lim ˝ Rel Rel !0

ˇ ˛˛ UCov .0; 1/ˇ˚Rel ˛ ˛ 0 Rel j UCov .0; 1/ j˚Rel

˛;

(6.30)

˛ ˛ is, as which is quite analogous to the nonrelativistic theorem (3.46). Here,j˚Rel before, the parent state (3.32), i.e., the limit of the corresponding target state, as the perturbation is adiabatically turned off, ˇ ˛˛ ˇ ˛ ˇ˚ D C ˛ lim ˇ ˛ .t/ ; (6.31) Rel Rel t !1

˛ ˛ ˛ is the (normalized) model (C ˛ is a normalization constant) andj0 ˛Rel D P jRel state. ˛ ˛ The state vector jRel satisfies a relativistic eigenvalue equation, analogous to the nonrelativistic (Schrödinger-like) Gell-Mann–Low equation (3.35), ˇ ˛˛ ˇ ˛ ˛ H0 C VF ˇRel D E ˛ˇRel ;

(6.32)

where VF is the perturbation, used in generating the evolution operator (6.5). In proving the relativistic form of the GML theorem, we observe that the covariant evolution operator differs from the corresponding nonrelativistic operator particularly by the replacement of the electron-field operators by the corresponding density operators (6.1). It then follows that the commutator of H0 with the covariant operator is the same as with the nonrelativistic operator, which implies that the proof in Sect. 3.3 can be used also in the covariant case.

132


A condition for the GML theorem to hold is as in the nonrelativistic case that the perturbation is time-independent in the Schrödinger picture (apart from damping), which is the case for the perturbation we shall use here (see further below).

6.5 Field-Theoretical Many-Body Hamiltonian In the unified MBPT–QED procedure, we shall apply the Coulomb gauge in order to be able to utilize the developments of the MBPT procedure. In this gauge, we separate the interaction between the electrons in the instantaneous Coulomb interaction and the transverse interaction, with the Coulomb part being (2.109) VC D

N X i <j

e2 : 4 0 rij

(6.33)

The exchange of a virtual transverse photon is represented by TWO perturbations of the one-body perturbation Z vT .t/ D

d3 x H.t; x/;

(6.34)

where the perturbation density given by (4.4) H.x/ D H.t; x/ D O .x/ ec˛ A .x/ O .x/

(6.35)

with A being the quantized, transverse radiation field (see Appendix F.2). The total perturbation is then VF D VC C vT : (6.36) The perturbation (6.35) represents the emission/absorption of a photon. Therefore, with this perturbation the GML equation works in a photonic Fock space,5 where the number of photons is not preserved. (The perturbation above is not time independent in the Schrödinger picture as required by the GML relation, but it can be transformed into equivalent, time-independent interactions, as will be demonstrated in Sect. 10.1). The model many-body Hamiltonian we shall apply is primarily a sum of Dirac single-electron Hamiltonians in an external (nuclear) field (Furry picture) (2.108) hD D c˛ pO C ˇmc 2 C vext :

5

(6.37)

Also, the Fock space is a form of Hilbert space, and therefore we shall refer to the Hilbert space with a constant number of photons as the restricted (Hilbert) space and the space with a variable number of photons as the (extended) photonic Fock space (see Appendix A.2).

6.5 Field-Theoretical Many-Body Hamiltonian

133

[As before, we may include an optional potential, u, in the model Hamiltonian (2.49) – and subtract the same quantity in the perturbation – in order to improve the convergence rate for many-electron systems.] However, since the number of photons is no longer constant in the space we work in, we have to include in the model Hamiltonian also the radiation field, HRad [see Appendix (G.12) and (B.20)], yielding H0 D

X

hD C HRad :

(6.38)

The full field-theoretical many-body Hamiltonian will then be H D H0 C VF D H0 C VC C vT

(6.39)

sometimes also referred to as the many-body Dirac Hamiltonian. This leads with the GML relation (6.32) to the corresponding Fock-space many-body equation6 H D E:

(6.40)

In comparing our many-body Dirac Hamiltonian with the Coulomb–Dirac–Breit Hamiltonian of standard MBPT (2.113), we see that we have included the radiation field, HRad , and replaced the instantaneous Breit interaction with the transverse field interaction, vT , in addition to removing the projection operators. Using second quantization (see Appendices B and E), the field-theoretical many-body Hamiltonian (6.39) becomes

Z

H D

d3 x O .x/ c ˛ pO C ˇmc 2 C vext .x/ ec˛ A .x/ O .x/ ZZ 1 e2 O .x2 / O .x1 /; CHRad C d3 x 1 d3 x 2 O .x1 / O .x2 / 2 4 0 r12 (6.41)

where vext .x/ is the external (nuclear) field of the electrons (Furry picture). We have here assumed that the Coulomb gauge is employed, and therefore the operator A .x/ represents only the transverse part of the radiation field. (As mentioned previously, it is quite possible to use the Coulomb gauge in QED calculation, as demonstrated by Adkins [1], Rosenberg [13], and others.)

6

This equation is not completely covariant, because it has a single time, in accordance with the established quantum-mechanical picture. This is the equal-time approximation, mentioned above and further discussed later. In addition, a complete covariant treatment would require that also the interaction between the electrons and the nucleus is treated in a covariant way by means of the exchange of virtual photons (see, for instance, [14]).

134


By treating the Coulomb and the transverse photon interactions separately,

a formal departure is made from a fully covariant treatment. However, this procedure is, when performed properly, in practice equivalent to the use of a covariant gauge. We define the wave-operator in analogy with the nonrelativistic case (2.37)7 j ˛ i D ˝j0˛ i

.˛ D 1 d /

(6.42)

but now acting in the extended photonic Fock space. The effective Hamiltonian has the same definition as before (2.38), which leads to Heff D PH ˝P

(6.43)

and the effective interaction is defined by Veff D Heff PH0 P D P .H H0 /˝P

(6.44)

or using the Hamiltonian (6.39) Veff D P VF ˝P D P VC C vT ˝P:

(6.45)

This is a Fock-space relation, and the corresponding relation in the restricted space without uncontracted photons is given by (6.123). By solving the many-body equation (6.40) iteratively, all possible perturbations will be produced. This is the basic principle of the covariant relativistic many-body perturbation procedure we shall develop in this book. How this can be accomplished will be discussed in the following. First, we shall treat the simple case of singlephoton exchange.

6.6 Green’s Operator 6.6.1 Definition The vacuum expectation used to define the Green’s function (5.8) contains singularities in the form of unlinked diagrams, where the disconnected parts represent the vacuum expectation of the S-matrix. This is a number, and it then follows that the singularities could be eliminated by dividing by this number. For the covariant

7

In the following, we shall leave out the subscript “Rel.”

6.6 Green’s Operator

135

evolution operator (CEO) (6.2), the situation is more complex, since this in an operator, and the disconnected parts will also in general be operators. Therefore, we shall here proceed in a somewhat different manner. As mentioned, we shall refer to the regular part of the CEO as the Green’s operator – in the

single-particle case denoted G.t; t0 / – due to its great similarity with the Green’s function. We define the single-particle Green’s operator by the relation8 U.t; t0 /P D G.t; t0 / P U.0; t0 /P;

(6.46)

where P is the projection operator for the model space, and analogously in the many-particle case. Below we shall demonstrate that the Green’s operator is regular. The definition of the Green’s operator contains the important concept of a heavy dot, which is defined in the following way. If the operators are disconnected, there is no difference between the dot product and an ordinary (normal-ordered) product. If the operators are connected to a diagram of ladder type in Fig. 6.3 (6.20), then we have seen that all interactions in an ordinary product depend on the energy of the initial state. We now introduce the convention that in a dot product the operators do not operate beyond the heavy dot.9 The definitions above imply that the interactions and the resolvents to the left of the dot depend on the energy of the unperturbed state at the position of the dot. If we operate to the right on the part of the model space PE of energy E and the intermediate model-space state lies in the part PE 0 of energy E 0 , we can express the two kinds of products as 8 ˆ < APE 0 BPE D A.E/ PE 0 B.E/PE ˆ :

(6.47) 0

A P BPE D A.E / P B.E/PE E0

E0

with the energy parameter of A equal to E in the first case and to E 0 in the second case. By the hooks we indicate that the operators must be connected by at least one contraction. We shall soon see the implication of this definition.

8

The Green’s operator is closely related – but not quite identical – to the reduced covariant evolution operator, previously introduced by the Gothenburg group [9]. 9 This can be compared with the situation in the MBPT Bloch equation (2.56), where – using the heavy dot – the folded term could be expressed ˝ P Veff P , indicating that the energy parameters of the wave operator depend on the intermediate model-space state.

136


6.6.2 Relation Between the Green’s Operator and Many-Body Perturbation Procedures From the conjecture (6.29) and the definition (6.46), we have in the limit of vanishing damping j ˛ .t/i D N˛ U.t; 1/j˚ ˛ i D N˛ G.t; 1/ P U.0; 1/P j˚ ˛ i ;

(6.48)

where N˛ is the normalization constant 1 N˛ D ˝ ˛ 0 jU.0; 1j˚ ˛ i

(6.49)

making the state vector intermediately normalized for t D 0. Here,j˚ ˛ i is the parent state (6.31), andj ˛ i D N˛ U.0; 1/j˚ ˛ i is the target state (for t D 0). The model state is j0˛ i D P j ˛ i D N˛ P U.0; 1/j˚ ˛ i : This leads directly to the relation

ˇ ˛ ˇ ˛ ˛ ˇ .t/ D G.t; 1/ˇ ˛ ; 0

(6.50)

which implies that the time dependence of the relativistic state vector is governed by the Green’s operator. Therefore, the Green’s operator can be regarded as a time-dependent wave operator – but it is NOT an evolution operator in the sense, discussed in Sect. 3.1. For the time t D 0, we have the covariant analogue of the standard wave operator of MBPT (2.37) j ˛ .0/i Dj ˛ i D ˝Covj ˛ i

(6.51)

˝Cov D G.0; 1/:

(6.52)

with

It follows directly from the definition (6.46) that P G.0; 1/P D P

(6.53)

and the relation above can also be expressed ˝Cov D 1 C QG.0; 1/:

(6.54)

6.6 Green’s Operator

137

We note here that it is important that the Green’s operator is defined with the dot product (6.46). The definition of the wave operator (2.37) can be expressed j ˛ i D ˝Cov P j ˛ i D ˝Cov P U.0; 1/j˚ ˛ i

(6.55)

indicating that the energy parameter of the wave operator depends on the intermediate model-space state. We shall also define a covariant effective interaction, analogous to the operator of MBPT (2.55). The time dependence of the relativistic state vector is formally the same as that of the nonrelativistic one (2.15) [which is verified below (6.120)] , i.e., in interaction picture ˇ ˛ ˛ ˇ ˇ ˛ ˛ ˇ .t/ D eit .E ˛ H0 /ˇ ˛ .0/ D eit .E ˛ H0 /ˇ ˛

(6.56)

or

@ ˇˇ ˛ ˛ (6.57) .t/ D .E ˛ H0 /j ˛ .t/i : @t With the relation (6.50), this yields for the time t D 0

@ ˇˇ ˛ ˛ @ i j ˛ i D i j0˛ i D .E ˛ H0 /j ˛ i : (6.58) .t/ G.t; 1/ @t @t t D0 t D0 i

Here, the r.h.s. becomes, using the GML relation (6.32) and the wave-operator relation (6.51), .H H0 /j ˛ i D VFj ˛ i D VF ˝Covj0˛ i : (6.59) These relations hold for all model states, which leads us to the important operator relation for the entire model space

@ G.t; 1/ i P D VF ˝Cov P; @t t D0

(6.60)

which we refer to as the reaction operator. Projecting this onto the model space, yields according to the definition (6.44) to the covariant relativistic effective interaction

@ Cov Veff D P VF ˝Cov P D P i G.t; 1/ P: t D0 @t

(6.61)

This is a relation in the photonic Fock space, closely analogous to the corresponding relation of standard MBPT (2.55) [c.f. the relation (6.123)]. Our procedure here is based upon quantum-field theory, and the Green’s operator can be regarded as a field-theoretical extension of the traditional wave-operator concept of MBPT, and it serves as a connection between field theory and MBPT.

138


6.7 Model-Space Contribution We shall now demonstrate how the singularities of the covariant evolution operator can be eliminated in the general multireference case. We assume that the initial time is t0 D 1. We also work in the equal-time approximation, where all final times are the same. We work in the restricted Hilbert space with no uncontracted photons and consider a ladder of complete single-photon interactions (6.21), transverse and Coulomb parts (see Fig. 6.3). (We shall later expand this to more general, irreducible interactions.) We start by expanding the relation (6.46) order-by-order, using the fact that U .0/ .0/P D P , U .0/ .t/P D G .0/ .t/ P U .0/ .0/P D G .0/ .t/P U .1/ .t/P D G .1/ .t/P C G .0/ .t/ P U .1/ .0/P U .2/ .t/P D G .2/ .t/P C G .1/ .t/ P U .1/ .0/P C G .0/ .t/ P U .2/ .0/P U .3/ .t/P D G .3/ .t/P C G .2/ .t/ P U .1/ .0/P C G .1/ .t/ P U .2/ .0/P C G .0/ .t/ P U .3/ .0/P;

(6.62)

etc. It follows from (6.21) that the time dependence of the ladder is given by eit .Ein Eout / , where Ein and Eout represent the incoming and outgoing energies. When operating on the part of the model space of energy E, the operator can be expressed as U.t/PE D eit .EH0 / U.0/PE : (6.63) Solving the equations (6.62) for the Green’s operator, we then have G .0/ .t/P D U .0/ .t/P G .1/ .t/P D U .1/ .t/P G .0/ .t/ P U .1/ .0/P G .2/ .t/P D U .2/ .t/P G .0/ .t/ P U .2/ .0/P G .1/ .t/ P U .1/ .0/P G .3/ .t/P D U .3/ .t/P G .0/ .t/ P U .3/ .0/P G .1/ .t/ P U .2/ .0/P G .2/ .t/ P U .1/ .0/P;

(6.64)

etc. We shall demonstrate that the negative terms above, referred to as counterterms, will remove the singularities of the evolution operator. It follows directly from the definition of the dot product above that the singularities due to disconnected parts are exactly eliminated by the counterterms. Therefore, we need only consider the connected (ladder) part, and we consider a fully contracted two-body diagram as an illustration (Fig. 6.3). It is sufficient for our present purpose to consider only positive intermediate states, as in (6.20).

6.7 Model-Space Contribution

139

6.7.1 Lowest Orders From the above, it follows that the zeroth-order Green’s operator is G .0/ .t; E/PE D U .0/ .t; E/PE D eit .EH0 / PE D PE :

(6.65)

(For clarity, we insert the energy parameter in the operator symbol.) In first order, we have from (6.64) G .1/ .t; E/PE D U .1/ .t; E/PE G .0/ .t; E 0 /PE 0 U .1/ .0; E/PE ;

(6.66)

where we observe that the Green’s operator in the counterterm has the energy parameter E 0 , due to the heavy dot in the expression (6.64). The first term is (quasi)singular, when the final state lies in the model space, and we shall show that this singularity is eliminated by the counterterm. From (6.63), we have PE 0 U .1/ .t; E/PE D PE 0 G .0/ .t; E/ U .1/ .0; E/PE (with the energy parameter E in the Green’s operator) and hence the corresponding part of the Green’s operator (6.66) becomes PE 0 G .1/ .t; E/PE D G .0/ .t; E/ G .0/ .t; E 0 / PE 0 U .1/ .0; E/PE

(6.67)

(P commutes with H0 ). According to (6.20) PE 0 U .1/ .0; E/PE D PE 0 .E/ Vsp .E/PE D PE 0

Vsp .E/ 1 Vsp .E/PE D PE 0 PE E H0 E E0 (6.68)

and hence we can express the first-order Green’s operator (6.66) as G .1/ .t; E/PE D QU .1/ .t; E/PE C

ıG .0/ .t; E 0 ; E/ PE 0 Vsp PE ; ıE

(6.69)

where QU .1/ .t; E/ D G .0/ .t; E/Q .E/V .E/PE . We assume here that there is a summation over E 0 , so that the entire intermediate model-space is covered. The difference ratio above is defined @G .0/ .t; E/ ıG .0/ .t; E 0 ; E/ G .0/ .t; E/ G .0/ .t; E 0 / ) D ; ıE E E0 @E

(6.70)

which turns into a derivative at complete degeneracy. Furthermore, .1/ PE PE 0 Vsp PE D PE 0 Veff

(6.71)

140


G .0/ .E /

6

6

U .1/ .E /

6 PE 0 -

6

6

6

PE

-

G .0/ .E 0 /

6

6

U .1/ .E /

6 PE 0 -

6

6

6

PE

)

6

6

6 PE 0 -

6

6

6

PE

Fig. 6.8 Illustration of the elimination of singularity of the first-order evolution operator, due to a final model-space state. The double bar represents the difference ratio/derivative of the zerothorder Green’s operator (c.f. Fig. 6.9)

is the first-order effective interaction, which is in accordance with the Fock-space expression (6.45) [see also (6.123)]. The first-order elimination process is illustrated in Fig. 6.8. In second order, we have from (6.21) .2/

.1/

.1/

U0 .t; E/Ladd PE D U0 .t; E/ U0 .0; E/ PE :

(6.72)

This can be (quasi)singular, if the final or intermediate state lies in the model space. If there is a model-space state only at the final state, the counterterm will lead – in complete analogy with the previous case (6.69) – to the contribution ıG .0/ .t; E 00 ; E/ PE 00 W0.2/ PE 0 ; ıE

(6.73)

where .2/

PE 00 W0 PE 0 D PE 00 Vsp .E/Q .E/Vsp .E/PE

(6.74)

is the second-order effective interaction without any intermediate model-space state. If there is an intermediate model-space state in the second-order evolution operator (6.72), we have U0.1/ .t; E/PE 0 U0.1/ .0; E/ PE D U0 .t; E/

PE 0 Vsp .E/ PE : E E0

(6.75)

The singularity will here be eliminated in a similar way by the corresponding counterterm (6.64). If also the final state lies in the model space, there is an additional singularity, which is eliminated by replacing U0.1/ .t; E/ by the corresponding Green’s operator (6.69), yielding for the entire second-order Green’s operator G .2/ .t; E/PE D G .0/ .t; E/Q .E/V E/Q .E/V E/PE C C

ıG .1/ .t; E 0 ; E/ PE 0 Vsp PE : ıE

ıG .0/ .t; E 00 ; E/ .2/ PE 00 W0 PE ıE (6.76)


6

6

G .1/ .E /

6 Q -

6

U .1/ .E /

6 PE 0 -

6

6

6

PE

141

-

6

6

6

6

G .1/ .E 0 /

6 Q -

6

6 Q -

6

U .1/ .E /

6 PE 0 -

6

6 PE 0 -

6

6

6

6

6

PE

)

PE

Fig. 6.9 Elimination of singularity of the second-order evolution operator, due to an intermediate model-space state. This leads to a residual contribution that corresponds to the folded diagram in standard many-body perturbation theory (Fig. 2.5). In addition, there can be a singularity at the final state, as in first order (see Fig. 6.8)

6

6

G .1/ .E /

6P C Q 6 -

U .1/ .E /

6P C Q 6 -

6

6

6

)

6

6P C Q 6 6 Q

6

6

6

-

6

+

6

6P C Q 6 6 P 6

6 6

Fig. 6.10 Elimination of the singularity of the second-order evolution operator due to an intermediate model-space state

The second-order elimination process, due to intermediate model-space state, is illustrated in Fig. 6.9, and the corresponding part of the Green’s operator is illustrated in Fig. 6.10. This process is quite analogous to the appearance of folded diagram, discussed in connection with standard MBPT (2.81). Since we are here dealing with Feynman diagrams, it is more logical to draw the “folded” part straight, indicating the position of the “fold” by a double bar from which the denominators of the upper part are to be evaluated. (The elimination process in first order has no analogy in standard MBPT, since there final model-space states do not appear in the wave function.) For t D 0, we have from (6.76) QG .2/ .0; E/PE D Q .E/Vsp .E/Q .E/Vsp .E/ PE C Q

ıG .1/ .t; E 0 ; E/ .1/ PE ; PE 0 Veff ıE (6.77)

which is quite analogous to the corresponding second-order wave operator in ordinary time-independent perturbation theory (2.69). The only difference is here that

142


the derivative of the first-order Green’s operator leads in addition to the standard folded term to a term with the energy-derivative of the interaction. The latter term is sometimes referred to as the reference-state contribution [12], but here we shall refer to both terms as the model-space contribution (MSC), which is more appropriate in the general multireference case. We have assumed so far that in the ladder the interactions are identical. If the interactions are different, some precaution is required. We see in the second-order expression that the differential/derivative in the last term should refer to the SECOND interaction, while if we treat this in an order-by-order fashion, we would get the differential of the FIRST interaction. If the interactions are in order V1 and V2 , then last term above becomes ı.Q V2 / PE 0 V1 PE ıE

(6.78)

(leaving out the arguments). This issue will be further discussed below.

6.7.2 All Orders The procedure performed above can be generalized to all orders of perturbation theory. We still consider a two-particle system in the ladder approximation. The treatment here follows mainly those of [10, 11] but is more general. We consider an evolution operator in the form of a ladder (6.21) with a general interaction, V .E/, and with all intermediate model-space states removed, including also the zeroth-order term, U0 .t; E/PE D G .0/ .t; E/ 1 C .E/V .E/ C .E/V .E/Q .E/V .E/ C PE ; (6.79) which may have a final model-space state. The corresponding Green’s operator is according to (6.64) U0 .t; E/PE G .0/ .t; E/ PE 0 .U0 .0; E/ 1/ PE ;

(6.80)

which can be expressed QU0 .t; E/PE C ıG .0/ .t; E 0 ; E/ PE 0 .U0 .0; E/ 1/ PE :

(6.81)

But in analogy with (6.68), we have PE 0 .U0 .0; E/ 1/ PE D

PE 0 W0 .E/PE ; E E0

(6.82)


143

where W0 is the effective interaction without intermediate model-space states or folds, in analogy with (6.74), W0 .E/PE D V .E/ C V .E/Q .E/V .E/ C PE : (6.83) Then the relation (6.81) becomes QU0 .t; E/PE C

ıG .0/ .t; E 0 ; E/ PE 0 W0 PE : ıE

(6.84)

The second term eliminates the singularity due to the final model-space state, and we shall refer also to this as a folded contribution, in analogy with those eliminating intermediate model-space singularities. The Green’s operator with no folds (intermediate or final) is G0 .t; E/PE D G .0/ .t; E/PE C QU0 .t; E/PE D G .0/ .t; E/ 1 C Q .E/V .E/ C Q .E/V .E/Q .E/V .E/ C PE : (6.85) The evolution operator with exactly one intermediate model-space state can be expressed (G .0/ .0/ D 1) QU0 .t; E/PE 0 .U0 .0; E/ 1/PE

(6.86)

and the folded part of (6.84) provides a final fold, yielding the Green’s operator with one intermediate or final fold, G1 .t; E/ D

ıG0 .t; E; E 0 / PE 0 W0 PE : ıE

(6.87)

The evolution operator with two intermediate folds can be expressed in analogy with (6.86) U2 .t; E/PE D QU0 .t; E/PE 0 .U0 .0; E/ 1/PE 0 .U0 .0; E/ 1/PE :

(6.88)

The two leftmost factors represent the Green’s operator (6.86) with one intermediate fold, and including also a final fold we can replace this by the operator (6.87) G1 .t; E 0 /PE 00 .U0 .0; E/ 1/PE : This represents the operator with exactly two intermediate or final model-space state with the singularity due to the leftmost one being eliminated. Eliminating also the second fold leads to the Green’s operator with two folds G2 .t; E/PE D

ıG1 .t; E 0 ; E/ PE 0 W0 .E/PE : ıE

(6.89)

144


Continuing this process leads to (with somewhat simplified notations) G.t; E/PE D G0 .t; E/ C G1 .t; E/ C G2 .t; E/ C PE " #

ıG0 .t; E/ ıG1 .t; E/ D G0 .t; E/ C C C W0 PE : ıE ıE This yields G.t; E/PE D G0 .t; E/PE C

ıG.t; E/ W0 PE : ıE

(6.90)

Here, the second term represents all intermediate/final folds (model-space contributions). This relation is valid for the entire model space and it is consistent with [11, Eq. (6.54)] but more general. The expressions given here are valid for all times and for the final state in P as well as Q spaces. The corresponding wave-operator relation is obtained by setting t D 0. We can find an alternative expression for the folded term in (6.90) by considering G D G0 C G1 C G2 C : From the expressions above, we find ıG0 W0 ıE

ıG1 ı ıG0 ı 2 G0 2 ıG0 ıW0 G2 D W C W0 D W0 W0 D W0 ıE ıE ıE ıE 2 0 ıE ıE ı 2 G0 2 ıG0 W1 D W C (6.91) ıE 2 0 ıE

G1 D

with Wn D

ıWn1 W0 ıE

being the effective interaction with exactly n folds. Similarly,10 10

ıG V ııG V 0 ıG GE GE 0 ı ıG ıE E E E E0 E D I V D ıE E E0 ıE ı E E E0 ıG VE ııG V C ııG V ııG V 0 ı2G ı G ıV E E0 E E E0 E E E0 E D ıE E D V C 0 E E ıE 2 ıE ıE ı VE V E 0 ı ıV 2 00 00 V D V E V E D VE DV : ıE ıE E E0 ıE

This can be generalized to

(see further [11, Appendix B]).

n X ı n .AB/ ı m A ı nm B D n ıE ı E m ı E nm mD0

(6.92)


145

ıG2 W0 ıE ı 3 G0 3 ı 2 G0 ıW0 2 ı 2 G0 ıG0 ıW1 D W C W1 W0 C W0 C W0 ıE 3 0 ıE 2 ıE ıE 2 ıE ıE

G3 D

or G3 D

ı 3 G0 3 ı 2 G0 ıG0 W C 2W1 W0 C W2 : ıE 3 0 ıE 2 ıE

Summing this sequence leads to ıG0 .W0 C W1 C W2 C / ıE 2 ı 3 G0 ı G0 .W03 C / C : C 2 .W02 C 2W0 W1 C / C ıE ıE 3

G D G0 C

(6.93)

It can be shown by induction [10] that this leads to G D G0 C

1 X n ı n G0 W0 C W1 C W2 C : n ıE

(6.94)

nD1

Here, Veff D W0 C W1 C W2 C

(6.95)

is the total effective interaction, which leads to G.t; E/PE D G0 .t; E/PE C

X ı n G0 .t; E/ n Veff PE : ıE n nD1

(6.96)

This relation is consistent with the results in [10, Eq. (100)] and [11, Eq. (61)], where more details of the derivations are given. As the previous relation (6.90), it is valid for all times and with the final state in Q as well as P space. In case the interactions are different, the derivatives should be taken of the latest interactions. We can generalize the treatment here and replace the single-photon potential by the two-body part of the complete irreducible multiphoton exchange potential (6.22) in Fig. 6.6, V ) V2 D V. It follows from the treatment here that the counterterms eliminate all singularities so that the Green’s operator is completely regular at all times.

146


6.7.2.1 Linkedness of the Green’s Operator All parts of the expansions above are linked, so this demonstrates that The Green’s operator is completely linked also in the multireference case. The linkedness of the single-particle Green’s operator can be expressed, us-

ing (6.4), " G .t; t0 / D 1

n Z ZZ Z 1 X i 1 d3 x d 3 x 0 d4 x1 d4 xn nŠ c

nD0

˝ ˇ

ˇ ˛ O H.x1 / H.xn / .x 0ˇT .x/ O 0 / ˇ0 e.jt1 jCjt2 j/

# linkedCfolded

(6.97) and similarly in the many-particle case. This represents a field-theoretical extension of the linked-diagram theorem of

standard many-body perturbation theory (2.82).

6.8 Bloch Equation for Green’s Operator* We now want to transform the general expression above for the Green’s operator into a general Bloch-type of equation (2.56) that, in principle, can be solved iteratively (self-consistently). Iterations can be performed, only if the in- and outgoing states contain only particle states of positive energy (no holes). Therefore, we assume this to be the case. If we have an interaction with hole states in or out, we can apply a Coulomb interaction, so that all in- and outgoing states are particle states, as will be discussed further in later chapters. We still work in the restricted Hilbert space with complete single-photon (or multiphoton) interactions. We want to have an equation of the form .n/

G ; H0 P D V G .n1/ P C folded (6.98) or

G .n/ PE D G .0/ PE C Q V G .n1/ C folded PE ;

(6.99)

where V is the last interaction. We start from the relation (6.90), G D G0 C

ıG W0 ; ıE

where G D G0 C G1 C G2 C

(6.100)

6.8 Bloch Equation for Green’s Operator

147

and Gm is the operator with exactly m intermediate/final folds, Gm D

ıGm1 W0 : ıE

Furthermore, the total effective interaction is (6.95) Veff D W0 C W1 C W2 C ;

(6.101)

where Wm is the effective interaction (6.92) with m folds and Wm D

ıWm1 W0 : ıE

(6.102)

The folded contribution of order n > 0 is according to (6.99) G .n/ Q V G .n1/ G .0/ D G0.n/ Q V G0.n1/ G .0/ C G1.n/ Q V G1.n1/ CG2.n/ Q V G2.n1/ C : We then see that in the case of no folds we have (6.85) G0.n/ Q V G0.n1/ G .0/ D 0:

(6.103)

In the case of a single fold, we have 1 D G1.n/ Q V G1.n1/ D

ıG

0

W0

ıE

.n/

Q V

ıG

0

ıE

W0

.n1/

:

Here, all terms cancel except those where the last factor of Q V is being differentiated in the first part of 1 and, in addition, terms with a fold in the final state. Obviously, those terms do not appear in the second part of the difference. This yields11 1 D

11

ı G

0

ıE

W0

.n/

;

(6.104)

Distinguishing the various interactions, we can write G0 D G .0/ 1 C Q V1 C Q V1 Q V2 C h ıG h ıG .0/ i ı.Q V1 / ı G0 i 0 1 D Q V 1 C G0 W0 D 1 C Q V2 C W0 ıE ıE ıE ıE ı 2 G0 i ı G1 h ı 2 G0 DW V W0 Q 1 ıE ıE 2 ıE 2 " ı 2 .Q V1 / ı 2 G .0/ ı G .0/ ı.Q V1 / 1 C Q V2 C C G .0/ 1 C Q V2 C D C 2 2 ıE ıE ıE ıE # ı.Q V1 / ı.Q V2 / ı G1 : 1 C Q V3 C C W0 DW C G .0/ ıE ıE ıE

148


where we have introduced the notation ı , with the asterisk indicating that the differentiation applies only to the last interaction, including the associated resolvent, Q V , ı .Q Va Q Vb / ı.Q Va / D Q V b ıE ıE

(6.105)

and, in addition, differentiation of G .0/ in case there is no Q V factor. In the case of two folds, we have

2 D

G2.n/

Q V

G2.n1/

.n/ .n1/ ıG1 ıG1 D Q V W0 W0 ıE ıE

.n/ ı ıG0 D W0 W0 ıE ıE

.n1/ ı ıG0 W 0 W0 Q V ıE ıE 2 2 .n/ .n1/ ı G0 ı G0 2 2 D .W0 / Q V .W0 / ıE 2 ıE 2 .n/ .n1/ ıG0 ıG0 W1 W1 C Q V : ıE ıE

With the convention above, we can express the folds

2 D

ı G0 W1 ıE

.n/

C

ı G1 W0 ıE

.n/ :

Continuing this process leads to the total folded contribution

ı G1 ı G0 C CC ıE ıE

.W0 C W1 C / D

ıG Veff ıE

with differentiation with respect to the last factor of Q V and to G .0/ , when no factor of Q V appears. We then have the generalized Bloch equation for an arbitrary energy-dependent

interaction (V) G D G .0/ C Q V G C where veff is given by (6.101).

ıG Veff ; ıE

(6.106)

6.8 Bloch Equation for Green’s Operator

149

This equation is valid also when the interactions are different, and then it can

be expressed more explicitly as (n > 0) G .n/ D Q Vn G .n1/ C

n1 X

ı G .m/ .nm/ Veff ; ıE mD0

(6.107)

where Vn is the last interaction and the operator G .m/ is formed by the m last interactions. We can check the formula (6.106) by considering the first few orders, G .0/ D eit .EH0 / ıG .0/ .1/ W0 ıE ıG .0/ .2/ ı G .1/ .1/ D Q V G .1/ C V C W0 ; ıE eff ıE

G .1/ D Q V G .0/ C G .2/

(6.108)

where the last term becomes ı G .1/ .1/ ı.Q V / .0/ .1/ ı 2 G .0/ .1/ 2 W0 : W0 D G W0 C ıE ıE ıE 2

(6.109)

This can easily be shown to reproduce the expansions (6.90) and (6.96). We can also illustrate the validity of the generalized Bloch equation (6.106) by considering the third-order case with different interactions, V1 ; V2 ; V3 , (see Fig. 6.11). For simplicity, we assume t D 0 and therefore make the replacement G ! ˝. If there is no model-space state directly after the first interaction (Fig. 6.11a), the contribution becomes

ı.Q V3 / .2/ ˝ Q V1 P D Q V3 Q V2 C P V2 Q V1 P ıE using the second-order expression (6.78) with the last two interactions (V2 ; V3 ).

a

6

6

V3

-

V2

Q -

V1

P

˝ .2/

b V3

-

V2

P -

V1

P

Fig. 6.11 Third-order Green’s operator with different interactions

ı˝ .2/ ıE .1/ Veff

150


With a model-space state directly after the first interaction (Fig. 6.11b), the contribution is

ı ı.Q V3 / ı˝ .2/ .1/ Q V3 Q V2 C Veff D P V 2 P V1 P ıE ıE ıE

ıQ V3 ı 2 Q V3 ıQ V2 D P V2 Q V2 C Q V3 C ıE ıE ıE 2 ıQ V3 ıV2 C P P P V1 P: ıE ıE We can now identify the terms above with the Bloch equation (6.106), where the differentiation should apply to the last interaction. Then we have ı.Q V2 / P V1 P ıE ıQ V3 ı 2 Q V3 Q V2 P V1 P C D P V2 P V1 P ıE ıE 2 ıQ V3 .2/ W0 C W1.2/ D ıE

ıQ V3 ıV2 D P V2 Q V1 P C P P V1 P ; ıE ıE

Q V3 ˝ .2/ D Q V3 Q V2 Q V1 P C Q V3 ı ˝ .2/ .1/ Veff ıE ı ˝ .1/ .2/ Veff ıE

the sum of which is identical to the sum of the two previous expressions. In most applications, we want to have an expression for the wave operator in the form of a Bloch equation, where we start from a wave operator ˝I and then add an interaction V that might be different from those involved in ˝I . The Bloch equation is then of the form ˝P D .˝I C Q .E/V ˝I C folded/P; where we want to find the form of the folded part. We then make the replacement ˝0 ) Q V ˝I0 in the expression (6.96), where ˝I0 is the wave operator without folds, yielding ˝D

X ı n .Q V ˝I0 / n Veff : ıE n nD0

(6.110)

The sum can be reformulated as, noting the modified differentiating rules given above, n X ı n .Q V ˝I0 / n X X ı m Q V m ı nm ˝I0 nm D V Veff Veff eff n ıE ıE m ıE nm nD0 nD0 mD0 X ı m Q V m X ı nm ˝I0 nm X ı m Q V m D D ˝I Veff ; V V eff eff m nm m ıE ıE ıE nDm mD0 mD0

˝D

(6.111)

6.9 Time Dependence of the Green’s Operator. Connection to the Bethe–Salpeter Equation

151

and this leads to the relation ˝ D Q V ˝ I C

X ı n .Q V / n ˝I Veff : n ıE nD1

(6.112)

Since the full wave operator appears only on the left-hand side, this equation does not have to be solved self-consistently. We can understand the appearance of the sequence of difference ratios above in the following way. Each model-space contribution (MSC) should contain a differentiation of all the following interactions. In ˝I , the last interaction, V, is not involved, and therefore a differentiation of Q V for each interaction in ˝I is required. We can illustrate the formula above with the third-order case considered previously (Fig. 6.11), now assuming that we have two Coulomb interactions (VC ), followed by an energy-dependent potential (V ). Then, we have instead ˝ .3/ P D Q V Q VC Q VC P C C

ıQ V P V C Q V C P ıE

ı 2 Q V ıQ V P VC P VC P; Q VC P VC P Q V Q2 VC P VC P C ıE ıE 2

which can be expressed ıQ V .1/ .1/ .2/ ˝I Veff C ˝I .0/ Veff ıE 2 ı 2 Q V .1/ .0/ V ˝ ; C I eff ıE 2

˝ .3/ P D Q V ˝I .2/ P C

where ˝I .1/ D Q VC

.1/

Veff D P VC P

˝I .2/ D Q VC Q VC Q2 VC P VC P

.2/

Veff D P VC Q VC P:

This is in agreement with the general formula (6.112).

6.9 Time Dependence of the Green’s Operator. Connection to the Bethe–Salpeter Equation 6.9.1 Single-Reference Model Space Operating with the relation (6.96) on a model function, 0 , of energy E0 , yields # " 1 n X ˇ ˛ ˇ ˛ ı G .t; E/ 0 ˇ 0 : G.t; E0 /ˇ0 D G0 .t; E/ C E/n (6.113) n ıE nD1 EDE0

152


We have here used the fact (6.44) that Veffj0 i D .E E0 /j0 i D Ej0 i :

(6.114)

The expansion (6.113) is a Taylor series, and the result can be expressed G.t; E0 /j0 i D G0 .t; E/j0 i ;

(6.115)

where G0 is the Green’s operator without model-space states (6.85). This implies that the sum in (6.113), representing the model-space contributions (MSC) to all orders, has the effect of shifting

the energy parameter from the model energy E0 to the target energy E . From the relations (6.21) and (6.64), we have the Green’s operator for the ladder without MSC in the present case, including also the zeroth order and the time factor, h G0 .t; E0 /j0 i D eit .E0 H0 / 1 C Q .E0 / V .E0 /

i C Q .E0 / V .E0 /Q .E0 / V .E0 / C j0 i :

(6.116)

The result (6.115) then implies that the Green’s operator with model-space contributions (MSC) becomes G.t; E0 /j0 i D G0 .t; E/j0 i D eit .E H0 / h iˇ ˛ 1 C Q .E/ V .E/ C Q .E/ V .E/Q .E/ V .E/ C ˇ0 : (6.117) shifting also the energy parameter of the time dependence.12

From this, it follows that i

ˇ ˛ ˇ ˛ @ G.t; E0 /ˇ0 D .E H0 / G.t; E0 /ˇ0 @t

(6.118)

and, using (6.52), @ i G.t; E0 / @t

12

!

ˇ ˛ ˇ ˛ ˇ0 D .E H0 / ˝ˇ0 :

(6.119)

t D0

We observe here that also the zeroth-order term has changed its time dependence, which is a consequence of the fact that the zeroth-order Green’s operator, G .0/ , is being modified by the expansion (6.96).


153

According to (6.50), the Green’s operator has the same time-dependence as the state vector, in the interaction picture j .t/i D eit .E H0 / j i

(6.120)

(withj i D j .0/i), which implies that the result above – which is a consequence of the initial conjecture (6.29) – is in accordance with the elementary quantummechanical result (2.15) and (3.2). Setting the time t D 0 yields with the identity (6.52), ˝j0 i D G.0; E0 /j0 i, the corresponding relation for the wave operator

ˇ ˛ j i D ˝j0 i D 1 C Q .E/ V .E/ C Q .E/ V .E/Q .E/ V .E/ C ˇ0 ; (6.121) which is the Brillouin–Wigner expansion of the wave function. From the relation (6.83), we have that the effective interaction without folds is ˇ ˛ ˇ ˛ W0 .E0 /ˇ0 D P V .E0 / C V .E0 /Q .E0 /V .E0 / C ˇ0 : (6.122) It can be shown in the same way as for the wave function that inclusion of the folds (MSC) leads to the replacement E0 ! E (see [10]) and to the expression for the full effective interaction (6.95) ˇ ˛ ˇ ˛ ˇ ˛ Veffˇ0 D W0 .E/ˇ0 D P V .E/ C V .E/Q .E0 /V .E/ C ˇ0 : But according to the definition (6.44), Veff D P .H H0 /˝P , which gives ˇ ˛ ˇ ˛ Veffˇ0 D P .H H0 /˝P D P V .E/˝ˇ0 :

(6.123)

This is an expression for the effective interaction in the restricted Hilbert space with no uncontracted photons, equivalent to the photonic-Fock-space relation (6.45). This is analogous to the MBPT result (2.55), but now the perturbation is energy dependent. We can generalize this treatment by replacing the single-photon potential V by the irreducible multiphoton potential in Fig. 6.6, V ) V2 D V. Then we have from (6.117) G.t; E0 /j0 i D G0 .t; E/j0 i D eit .E H0 /

ˇ ˛ 1 C Q .E/ V.E/ C Q .E/ V.E/Q .E/ V.E/ C ˇ0 (6.124) and ˇ ˛ ˇ ˛ ˇ ˛ Veffˇ0 D P .E H0 / ˝ˇ0 D P V.E/˝ˇ0 :

(6.125)

154


From the relation, (6.118) we have

ˇ ˛ ˇ ˛ @ ˇ0 D Q.E H0 / ˝ˇ0 Q i G.t; E0 / @t t D0

ˇ ˛ D Q V.E/ C V.E/Q .E0 / V.E/ C ˇ0 ˇ ˛ D QV.E/ ˝ˇ0 : Combining this with (6.125) leads to the Schrödinger-like equation in the restricted space .H0 C V.E//j i D Ej i

(6.126)

and an energy-dependent Hamilton operator H D H0 C V.E/:

(6.127)

These relation can be compared with the corresponding GML relations (6.32) and (6.39) in the photonic Fock space. The equation (6.126) is identical to the effectivepotential form of the Bethe–Salpeter equation (9.20).

6.9.2 Multireference Model Space We shall now investigate the time dependence of the Green’s operator in a general, quasi-degenerate model space. We can express the relation (6.96), using the general perturbation, as G.t; E/PE D G0 .t; E/PE C

1 X ı n G0 .t; E/ n Veff PE ; ıE n nD1

(6.128)

valid in the general multireference (quasi-degenerate) case, where PE is the part of the model space with energy E and Veff is given by (6.138). This can be formally expressed as an operator relation G.t; H0 /P D G0 .t; H0 /P C

1 X ı n G0 .t; H0 / n Veff P; ı.H0 /n nD1

(6.129)

valid in the entire model space. We have here introduced the symbol A , which implies that the operator A operates directly on the model-space˛ state to the right. ˛ Bj0˛ D E ˛ Bj0˛ D Thus, H0˛BPE D EBPE D BH0 PE . Similarly, Heff BHeff j0˛ .


155

In analogy with (6.45), we have Veffj0˛ i D P .E ˛ H0 /˝j0˛ i D P V.E ˛ /˝j0˛ i

(6.130)

or in operator form Veff P D P .Heff H0 /˝ P D P V.Heff /˝P:

(6.131)

The relation (6.129) is a Taylor expansion in analogy with (6.115), yielding G.t; H0 / P D G0 .t; Heff /P

(6.132)

Heff D PH0 P C Veff :

(6.133)

using the fact (6.44) that From (6.116), it follows

G0 .t; E/PE D eit .EH0 / 1 C Q .E/ V.E/ C Q .E/ V.E/Q .E/ V.E/ C PE (6.134) or in operator form G0 .t; H0 /P D eit .H0 H0 / 1 C Q .H0 / V.H0 /

CQ .H0 / V.H0 /Q .H0 / V.H0 / C P:

(6.135)

This leads in analogy with (6.117), using the relation (6.115), to G.t; H0 /P D G0 .t; Heff /P D eit .Heff H0 / 1 C Q .Heff / V.Heff /

CQ .Heff / V.Heff / Q .Heff / V.Heff / C P: (6.136) From this, we conclude that the general time dependence of the Green’s operator is given by

i

@ H0 / G.t; H0 /P: G.t; H0 /P D .Heff @t

(6.137)

This gives with (6.133) P

@ i G.t; H0 / P D Veff P; @t t D0

which is the expected result.

(6.138)

156


In analogy with the single-reference case, the effective interaction becomes Veff D P V.Heff /˝ P

(6.139)

and the Schrödinger-like equation [10, Eq. (113)] .H0 C V.E ˛ //j ˛ i D E ˛ /j ˛ i :

(6.140)

This agrees with the equation derived in [10, Eq. (133)] and it is equivalent to the Bethe–Salpeter–Bloch equation, discussed in Chap. 9 (9.30).

References 1. Adkins, G.: One-loop renormalization of Coulomb-gauge QED. Phys. Rev. D 27, 1814–20 (1983) ˚ en, B.: QED effects in excited states of helium-like ions. Ph.D. thesis, Department of 2. As´ Physics, Chalmers University of Technology and University of Gothenburg, Gothenburg, Sweden (2002) 3. Bjorken, J.D., Drell, S.D.: Relativistic Quantum Fields. Mc-Graw-Hill Pbl. Co, N.Y. (196) 4. Bjorken, J.D., Drell, S.D.: Relativistic Quantum Mechanics. Mc-Graw-Hill Pbl. Co, N.Y. (1964) 5. E.Lindroth, M˚artensson-Pendrill, A.M.: Isotope Shifts and Energies of the 1s2p States in Helium. Z. Phys. A 316, 265–273 (1984) 6. Feynman, R.P.: The Theory of Positrons. Phys. Rev. 76, 749–59 (1949) 7. Lindgren, I.: Can MBPT and QED be merged in a systematic way? Mol. Phys. 98, 1159–1174 (2000) ˚ en, B., Salomonson, S., M˚artensson-Pendrill, A.M.: QED procedure applied 8. Lindgren, I., As´ to the quasidegenerate fine-structure levels of He-like ions. Phys. Rev. A 64, 062,505 (2001) ˚ en, B.: The covariant-evolution-operator method in bound9. Lindgren, I., Salomonson, S., As´ state QED. Physics Reports 389, 161–261 (2004) 10. Lindgren, I., Salomonson, S., Hedendahl, D.: Many-body-QED perturbation theory: Connection to the two-electron Bethe-Salpeter equation. Einstein centennial review paper. Can. J. Phys. 83, 183–218 (2005) 11. Lindgren, I., Salomonson, S., Hedendahl, D.: Many-body procedure for energy-dependent perturbation: Merging many-body perturbation theory with QED. Phys. Rev. A 73, 062,502 (2006) 12. Mohr, P.J., Plunien, G., Soff, G.: QED corrections in heavy atoms. Physics Reports 293, 227–372 (1998) 13. Rosenberg, L.: Virtual-pair effects in atomic structure theory. Phys. Rev. A 39, 4377–86 (1989) 14. Shabaev, V.M., Artemyev, A.N., Beier, T., Plunien, G., Yerokhin, V.A., Soff, G.: Recoil correction to the ground-state energy of hydrogenlike atoms. Phys. Rev. A 57, 4235–39 (1998) 15. Stuckelberg, E.C.G.: . Helv.Phys.Acta 15, 23 (1942) 16. Uehling, E.A.: Polarization Effects in the Positron Theory. Phys. Rev. 48, 55–63 (1935) 17. Wichmann, E.H., Kroll, N.M.: Vacuum Polarization in a Strong Coulomb Field. Phys. Rev. 101, 843–59 (1956)

Chapter 7

Numerical Illustrations to Part II

In this chapter, we shall give some numerical illustrations of the three QED methods described in Part II, the S-matrix, the two-times Green’s function, and the covariantevolution-operator methods.

7.1 S-Matrix 7.1.1 Electron Self-Energy of Hydrogen-Like Ions In the early days of quantum-electrodynamics, the effects were calculated analytically, applying a double expansion in ˛ and Z˛. For high nuclear charge, Z, such an expansion does not work well, and it is preferable to perform the evaluation numerically to all orders of Z˛. The first numerical evaluations of the electron self-energy on heavy, many-electron atoms were performed by Brown et al. in late 1950s [14] and by Desiderio and Johnson in 1971 [18], applying a scheme devised by Brown, Langer, and Schaefer [13] (see Sect. 12.3). An improved method for self-energy calculations, applicable also for lighter systems, was developed and successfully applied to hydrogen-like ions by Mohr [34, 39–42]. The energy shift due to the first-order electron self-energy is conventionally expressed as: ˛ .Z˛/4 E D F .Z˛/ mc 2 (7.1) n3 where n is the main quantum number. The function F .Z˛/ is evaluated numerically, and some results are given in Table 7.1. Performing accurate self-energy calculations for low Z is complicated due to slow convergence. Mohr has estimated the first-order Lamb shift (self-energy C vacuum polarization) by means of elaborate extrapolation from heavier elements and obtained the value 1057.864(14) MHz for the 2s 2p1=2 shift in neutral hydrogen [40], in excellent agreement with the best experimental value at the time, 1057.893(20) MHz. More recently, Jentschura et al. [26] have extended the method


157

158 Table 7.1 The F .Z˛/ function for the ground state of hydrogen-like mercury

Table 7.2 Ground-state Lamb shift of hydrogen-like uranium (in eV, mainly from [43])

7 Numerical Illustrations to Part II Reference Desiderio and Johnson [18] Mohr [38] Blundell and Snyderman [9] Mohr [34] Correction Nuclear size First-order self-energy Vacuum polarization Second-order effects Nuclear recoil Nuclear polarization

F .Z˛/ 1.48 1.5032(6) 1.5031(3) 1.5027775(4)

Value 198:82 355:05 88:59 1:57 0:46 0:20

Total theory

463:95

Experimental

460:2.4:6/

Reference [34, 45] [49]

of Mohr to calculate directly the self-energy of light elements down to hydrogen with extremely high accuracy. Accurate calculations have also been performed for highly excited states [27]. The original method of Mohr was limited to point-like nuclei but was extended to finite nuclei in a work with Gerhard Soff [45]. An alternative method also applicable to finite nuclei has been devised by Blundell and Snyderman [9, 10].

7.1.2 Lamb Shift of Hydrogen-Like Uranium In high-energy accelerators, like that at GSI in Darmstadt, Germany, highly charged ions up to hydrogen-like uranium can be produced. For such systems the QED effects are quite large, and accurate comparison between experimental and theoretical results can here serve as an important test of the QED theory in extremely strong electromagnetic fields – a test that has never been performed before. The first experimental determination of the Lamb-shift in hydrogen-like uranium was made by the GSI group (Stöhlker, Mokler et al.) in 1993 [56]. The result was 429(23) eV, a result that has gradually been improved by the group, and the most recent value is 460.2(4.6) eV [55]. The shift is here defined as the experimental binding energy compared to the Dirac theory for a point nucleus, implying that it includes also the effect of the finite nuclear size. In Table 7.2, we show the various contributions to the theoretical value. The self-energy contribution was evaluated by Mohr [34] and the finite-nuclear-size effect by Mohr and Soff [45]. The vacuum polarization, including the Wickmann–Kroll correction (see Sect. 4.6.3), was evaluated by Persson et al. [49]. The second-order QED effects, represented by the diagrams in Fig. 7.1, have also been evaluated. Most of the reducible part was evaluated by Persson et al. [48]. The last two irreducible too-loop diagrams are much more elaborate to calculate and have only recently been fully evaluated by Yerokhin et al. [60].1 1

See Sect. 2.6.

7.1 S-Matrix

159

-

-

-

6

6

6

- 6 6

66

6

6 1 ? 6

6

6

6

6

6

6

6

66

6

6

6

6

6

6

6 6

6

Fig. 7.1 Second-order contributions to the Lamb shift of hydrogen-like ions (c.f. Fig. 5.3)

The main uncertainty of the theoretical calculation on hydrogen-like uranium stems from the finite-nuclear-size effect, which represents almost half of the entire shift from the Dirac point-nuclear value. Even if the experimental accuracy would be significantly improved, it will hardly be possible to test with any reasonable accuracy the second-order QED effects, which are only about 1% of the nuclear-size effect. For that reason, other systems, like lithium-like ions, seem more promising for testing such effects.

7.1.3 Lamb Shift of Lithium-Like Uranium The 2s 2p1=2 Lamb shift of lithium-like uranium was measured at the Berkeley HILAC accelerator by Schweppe et al. in 1991 [54]. The first theoretical evaluations of the self-energy was performed by Cheng et al. [15] and the complete first-order shift, including vacuum polarization, by Blundell [7], Lindgren et al. [31],

160

7 Numerical Illustrations to Part II Table 7.3 2s 2p1=2 Lamb shift of lithium-like uranium (in eV) Correction Ref. [7] Ref. [48] Relativistic MBPT 322.41 322.32 First-order self-energy 53:94 54:32 First-order vacuum polarization (12.56) 12.56 First-order self-energy C vac. pol. 41:38 41:76 Second-order self-energy C vac. pol. 0.03 Nuclear recoil (0.10) (0:08) Nuclear polarization (0.10) (0.03) Total theory

280.83(10)

Experimental

280.59(9)

Fig. 7.2 Feynman diagrams representing the two-photon exchange (ladder and cross) for helium-like ions

6 6 6

280.54(15)

-

Ref. [58] 322.10

41:77 0.17 0:07 0:07 280.48(20)

6

6

6

6

6 6 6

6

6

6

and Persson et al. [48], the latter calculation including also some reducible secondorder QED effects. Later, more complete calculations were performed by Yerokhin et al. [58]. The results are summarized in Table 7.3. In lithium-like systems, the nuclear-size effect is considerably smaller than in the corresponding hydrogen-like system and can be more easily accounted for. The second-order QED effects in Li-like uranium are of the same order as the present uncertainties in theory and experiment, and with some improvement these effects can be tested. Therefore, systems of this kind seem to have the potential for the most accurate test of high-field QED at the moment.

7.1.4 Two-Photon Nonradiative Exchange in Helium-Like Ions Accurate S-matrix calculations of the nonradiative two-photon exchange for helium-like ions (ladder and cross), corresponding to the Feynman diagrams in Fig. 7.2, have been performed by Blundell et al. [8] and Lindgren et al. [30]. The results are illustrated in Fig. 7.3 (taken from [30]). In the figure, the contributions are displayed versus the nuclear charge, relative to the zeroth-order nonrelativistic ionization energy, Z 2 =2 (in atomic Hartree units). The vertical scale is logarithmic, so that 1 corresponds to ˛, 2 to ˛ 2 , etc. As comparison, we show in the top picture of Fig. 7.3 the energy contribution due to first-order Coulomb and Breit interactions as well as the first-order Lamb shift, corresponding to Feynman diagrams shown in the top line of Fig. 7.4. For low Z, the first-order Coulomb interaction is proportional to Z, the firstorder Breit interaction to Z 3 ˛ 2 , and the first-order Lamb shift to Z 4 ˛ 3 . For high

7.1 S-Matrix

161

Fig. 7.3 Various contributions to the ground-state energy of He-like ions. The top picture represents the first-order contributions, the middle picture the second-order contributions in the NVPA as well as the screened Lamb shift, and the bottom picture contributions due to retardation and virtual pairs (see Fig. 7.4). The values are normalized to the nonrelativistic ionization energy, and the scale is logarithmic (powers of the fine-structure constant ˛)

162

6 66

7 Numerical Illustrations to Part II

6

6

6

6

6

6

6

6

6

6

6

6

6

6

6

6

6

6

6

6

6 6 6

6 6 6

6

6 6

6

6

6

6

66

6

6 6 6

6 6

6

6 6

6

6

6

6

6

6

6 6

6 6

6

6

6 6

6

6 6 6 6

6

6 6 6

-

6 6 6

Fig. 7.4 Feynman diagrams representing the one- and two-photon exchange, separated into Coulomb, instantaneous Breit and retarded Breit interactions

Z, we can replace Z˛ by unity, and then after dividing by Z 2 , all first-order effects tend to ˛ as Z increases, as is clearly seen in the top picture of Fig. 7.3 (see also Fig. 10.9 and Table 10.1). An additional Coulomb interaction reduces the effect for small Z by a factor of Z. Therefore, the Coulomb–Coulomb interaction, i.e., the leading electron correlation, is in first order independent of Z and the Coulomb–Breit interaction proportional to Z 2 ˛ 2 . The screened Lamb shift is proportional to Z 3 ˛ 3 and the second-order Breit interaction (in the no-pair approximation) to Z 4 ˛ 4 . After division with Z 2 , we see (second picture of Fig. 7.3) that all second-order effects tend to ˛2 . The corresponding Feynman diagrams are shown in the second row of Fig. 7.4. The third picture in Fig. 7.3 shows the effect of the retarded Coulomb–Breit and Breit–Breit interactions without and with virtual pairs, corresponding to diagrams in the bottom row of Fig. 7.4. For low Z, these effects are one order of ˛ smaller than the corresponding unretarded interactions with no virtual pairs, while for high Z they tend – rather slowly – to the same ˛2 limit. It is notable that for the Coulomb–Breit interactions the retardation and virtual pairs have nearly the same effect but with opposite sign. For the Breit–Breit interactions, the effects of single and double pairs have opposite sign and the total effect changes its sign around Z D 40.

7.1 S-Matrix Table 7.4 Two-photon effects on some excited states of helium-like ions (in Hartree, from [44])

Table 7.5 Two-electron effects on ions (in eV, from [50]) Z Plante et al. Indelicato et al. 32 652.0 562.1 54 1,028.4 1,028.2 66 1,372.2 1,336.5 74 1,574.8 1,573.6 83 1,880.8

163 Z 30

MBPT QED 50 MBPT QED 80 MBPT QED

2 3 S1 49 541 8:7 53 762 64 66 954 966

2 3 P0 88 752 145 123 159 1340 251 982 9586

2 3 P2 75 352 77.6 79 949 767 93 5482

the ground-state energy of helium-like Drake 562.1 1,028.8 1,338.2 1,576.6 1,886.3

Persson et al. 562.0 1,028.2 1,336.6 1,573.9 1,881.5

Expt’l 562.5˙1:5 1,027.2˙3:5 1,341.6˙4:3 1,568˙15 1,876˙14

More recently, Mohr and Sapirstein have performed S-matrix calculations also on the excited states of helium-like ions and compared with second-order MBPT calculations to determine the effect of nonradiative QED, retardation, and virtual pairs [44], and some results are shown in Table 7.4.

7.1.5 Electron Correlation and QED Calculations on Ground States of Helium-Like Ions The two-electron effect on the ground-state energy of some helium-like ions has been measured by Marrs et al. at Livermore Nat. Lab. by comparing the ionization energies of the corresponding helium-like and hydrogen-like ions [33]. (The larger effect due to single-electron Lamb shift is eliminated in this type of experiment.) Persson et al. [50] have calculated the two-electron contribution by adding to the all-order MBPT result the effect of two-photon QED, using dimensional regularization (see Chap. 12). The results are compared with the experimental results as well as with other theoretical estimates in Table 7.5. The results of Drake were obtained by expanding relativistic and QED effects in powers of ˛ and Z˛, using Hylleraas type of wave functions [20]. The calculations of Plante et al. were made by means of relativistic MBPT and adding first-order QED corrections taken from the work of Drake [52], and the calculations of Indelicato et al. were made by means of multiconfigurational Dirac–Fock with an estimate of the Lamb shift [25]. The agreement between experiments and theory is quite good, although the experimental accuracy is not good enough to test the QED parts, which lie in the range 1–5 eV. The agreement between the various theoretical results is very good – only the results of Drake are somewhat off for the heaviest elements, which is due to the shortcoming of the power expansion.

164


7.1.6 g-Factor of Hydrogen-Like Ions: Mass of the Free Electron The Zeeman splitting of hydrogen-like ions in a magnetic field is another good test of QED effects in highly charged ions. The lowest-order contributions to this effect are represented by the Feynman diagrams in Fig. 7.5. The bound-electron g-factor can be expanded as [51]:

i ˛ p 1h 1 C 2 1 .Z˛/2 C gJ D 2 3

1 .Z˛/2 C C 2 12

;

(7.2)

where Z is the nuclear charge. The first term represents the relativistic value with a correction from the Dirac value of order ˛ 2 . The second term, proportional to ˛, is the leading QED correction, known as the Schwinger correction, and the following term, proportional to ˛ 3 , is the next-order QED correction, first evaluated by Grotsch [23]. Numerical calculations to all orders in Z˛ have been performed by Blundell et al. [11] [only self-energy part, (b,c) in Fig. 7.5] and by the Gothenburg group [6, 51] [incl. the vacuum polarization (d,e)]. The results are displayed in Fig. 7.6, showing the comparison between the Grotsch term (the leading QED correction beyond the Schwinger correction) and the numerical result. (The common factor of 2˛= has been left out.) More accurate calculations have later been performed by the St Petersburg group, including also two-loop corrections and the nuclear recoil [59, 61]. The g-factors of hydrogen-like ions have been measured with high accuracy by the Mainz group, using an ion trap of Penning type [4, 24]. The accuracy of the experimental and theoretical determinations is so high that the main uncertainty is due to the experimental mass of the electron. Some accurate dates for H-like carbon are shown in Table 7.6. By fitting the theoretical and experimental values, a value of the electron mass (in atomic mass units) me D 0:0005485799093.3/ is deduced from the carbon experiment and the value me D 0:0005485799092.5/ from a similar experiment on oxygen [4]. These results are four times more accurate than the previously accepted value, me D 0:0005485799110.12/ [37]. The new value is now included in the latest adjustments of the fundamental constants [35, 36].

a 6 6

c

b 6

6

6 6 6

66

6

e

d

6 6

6 6 6

?

6 6

Fig. 7.5 Feynman diagrams representing the lowest-order contributions to the Zeeman effect of hydrogen-like ions. Diagrams (b) and (c) represent the leading self-energy correction to the firstorder effect (a, d and e) the leading vacuum-polarization correction

7.2 Green’s-Function and Covariant-Evolution-Operator Methods

165

Fig. 7.6 The first-order, numerically evaluated, QED correction to the gj value of hydrogen-like ions, compared with the leading analytical (Grotsch) term (7.2). Both results are first-order in ˛ but the numerical result is all order in Z˛, while the Grotsch result contains only the leading term (from [51]). A common factor 2˛= is left out

Table 7.6 Theoretical contributions to the g-factor of hydrogen-like carbon (mainly from [5])

Correction Dirac theory Finite nuclear-size corr. Nuclear recoil Free-electron QED, first order Free-electron QED, higher orders Bound-electron QED, first order Bound-electron QED, higher orders Total theory

Value 1.998 721 3544 C0.000 000 0004 C0.000 000 0876 C0.002 322 8195 C0.000 003 5151 C0.000 000 8442 0.000 000 0011 2.001 041 5899

7.2 Green’s-Function and Covariant-Evolution-Operator Methods 7.2.1 Fine-Structure of Helium-Like Ions The two-times Green’s function and the covariant-evolution-operator methods have the important advantage over the S-matrix formulation that they can be applied also to quasi-degenerate energy levels. As an illustration, we consider here the evaluation of some fine-structure separations of the lowest P state of helium-like ions (see Table 7.7). The calculations of Plante et al. [52] are relativistic many-body calculations in the NVPA scheme (see Sect. 2.6) with first-order QED-energy cor˚ en et al. [2, 29], rections, taken from the work of Drake [20]. The calculations by As´

166


Table 7.7 The 1s2p 3 P fine structure of He-like ions. (Values for Z D 2, 3 given in MHz and the remaining ones in Hartree) 3 3 Z 3 P1 3 P0 P2 3 P0 P2 3 P1 Ref. Expt’l Ref. Theory 2 29616.95166(70) 2291.17759(51) Gabrielse et al. [62] 29616.9527(10) Giusfredi et al. [22] 29616.9509(9) Hessels et al. [21] 2291.17753(35) Hessels et al. [12] 29616.9523(17) 2291.1789(17) Pachucki et al. [57] 3

155704.27(66) 155703.4(1,5)

9

62678:41.65/ 62678:46.98/ 62679:4.5/

Riis et al. [53] Clarke et al. [16] Drake et al. [20]

701(10) 680 681 690

5064(8) 5050 5045 5050

4364.517(6) 4362(5) 4364 4364

Myers et al. [46]

10 1371(7) 1361(6) 1370 1370

8458(2) 8455(6) 8469 8460

7087(8) 7094(8) 7099 7090

Curdt et al. [17]

12 3789(26) 3796(7) 3778(10) 3796 3800,1

20069(9)

16280(27)

Curdt et al. [17] Myers et al. [47]

20046(10) 20072 20071

16268(13) 16276

14 8108(23) 8094 18 23692 23790

40707(9) 40708(23) 40707 40712

Drake et al. [20] Plante et al. [52] ˚ en et al. [2, 32] As´ Drake et al. [20] Plante et al. [52] ˚ en et al. [29] As´

Drake et al. [20] Plante et al. [52] Artemyev et al. [1] Curdt et al. [17]

32601(33) 32613

124960(30) 124810(60) 124942 101250 124940 101150 124945(3)

Drake et al. [20] Plante et al. [52] Artemyev et al. [1] Kukla et al. [28] Drake et al. [20] Plante et al. [52] ˚ en et al. [29] As´ Artemyev et al. [1]

using the recently developed covariant-evolution-operator method, was the first numerical evaluation of QED effects (nonradiative) on quasi-degenerate energy levels. It can be noted that the energy of the 1s2p 3 P1 state, which a linear combination of the closely spaced states 1s2p1=2 and 1s2p1=3 , could not be evaluated by the S-matrix formulation (see, for instance, the above-mentioned work of Mohr and Sapirstein [44]). Later, calculations have also been performed on these systems by the St Petersburg group, using the two-times-Green’s-function method [1], where also the radiative parts are evaluated numerically. The accuracy of the experimental and theoretical fine-structure results is not sufficient to distinguish between the first-order energy QED corrections and the

7.2 Green’s-Function and Covariant-Evolution-Operator Methods Table 7.8 The transition 1s2s 1 S0 1s2p 3 P1 for He-like Si (in cm1 )

Table 7.9 Two-photon calculations on the 1s2s 1 S; 3 S states of helium-like ions (in Hartree, ˚ en first two columns from As´ et al. [3], last column from Mohr and Sapirstein [44])

Expt’l Theory

Z 10 18 30 60

MBPT QED MBPT QED MBPT QED MBPT QED

7230.585(6) 7231.1 7229(2)

167 Reference Myers et al. [19] Plante et al. [52] Artemyev et al. [1]

2 3 S0

2 3 S1

2 3 S1

116 005 6.2 119 381 3.8 128 349 93 177 732 2358

47 638 1:2 48 158 4.6 49 542 49 541 6.9 8.7 57 025 57 023 216 224

˚ numerical evaluation of Asen and Artemyev. On the other hand, the experimental accuracy of the separation of Fluorine (Z D 9) seems to be sufficient to test even higher-order QED effects. Here, present theory cannot match the experimental accuracy, but this might be a good testing case for the new combined QED-correlation procedure, discussed in the following. As a second illustration, we consider the transition 1s2s 1 S0 1s2p 3 P1 for He-like silicon, which has recently been very accurately measured by Myers et al. [19] (see Table 7.8). Corresponding calculations have been performed by Plante et al. [52], using relativistic MBPT with first-order QED correction, and by Artemyev et al. [1], using the two-times Green’s function. Here, it can be seen that the experiment is at least two orders of magnitude more accurate than the theoretical estimates. Also here the combined MBPT–QED corrections are expected to be significant.

7.2.2 Energy Calculations of 1s2 s Levels of Helium-Like Ions ˚ en et al. [2, 3] The covariant-evolution-operator method has also been applied by A´ to evaluate the two-photon diagrams in Fig. 7.2 for the first excited S states of some helium-like ions. The results are compared with relativistic MBPT results, to determine the nonradiative QED effects, as in Table 7.4 above. The results are shown in Table 7.9, where comparison is also made with some results of Mohr and Sapirstein [44].

168


References 1. Artemyev, A.N., Shabaev, V.M., Yerokhin, V.A., Plunien, G., Soff, G.: QED calculations of the nD1 and nD2 energy levels in He-like ions. Phys. Rev. A 71, 062,104 (2005) ˚ en, B.: QED effects in excited states of helium-like ions. Ph.D. thesis, Department of 2. As´ Physics, Chalmers University of Technology and University of Gothenburg, Gothenburg, Sweden (2002) ˚ en, B., Salomonson, S., Lindgren, I.: Two-photon exchange QED effects in the 1s2s 1 S and 3. As´ 3 S states of heliumlike ions. Phys. Rev. A 65, 032,516 (2002) 4. Beier, T., Djekic, S., Häffner, H., Indelicato, P., Kluge, H.J., Quint, W., Shabaev, V.S., Verdu, J., Valenzuela, T., Werth, G., Yerokhin, V.A.: Determination of electron’s mass from g-factor experiments on 12 C C5 and 16 O C7 . Nucl. Instrum. Methods 205, 15–19 (2003) 5. Beier, T., Häffner, H., Hermanspahn, N., Karshenboim, S.G., Kluge, H.J.: New Determination of the Electron’s Mass. Phys. Rev. Lett. 88, 011,603.1–4 (2002) 6. Beier, T., Lindgren, I., Persson, H., Salomonson, S., Sunnergren, P., Häffner, H., Hermanspahn, N.: gj factor of an electron bound in a hydrogenlike ion. Phys. Rev. A 62, 032,510 (2000) 7. Blundell, S.: Calculations of the screened self-energy and vacuum polarization in Li-like, Nalike, and Cu-like ions. Phys. Rev. A 47, 1790–1803 (1993) 8. Blundell, S., Mohr, P.J., Johnson, W.R., Sapirstein, J.: Evaluation of two-photon exchange graphs for highly charged heliumlike ions. Phys. Rev. A 48, 2615–26 (1993) 9. Blundell, S., Snyderman, N.J.: Basis-set approach to calculating the radiative self-energy in highly ionized atoms. Phys. Rev. A 44, R1427–30 (1991) 10. Blundell, S.A.: Accurate screened QED calculations in high-Z many-electron ions. Phys. Rev. A 46, 3762–75 (1992) 11. Blundell, S.A., Cheng, K.T., Sapirstein, J.: Radiative corrections in atomic physics in the presence of perturbing potentials. Phys. Rev. A 55, 1857–65 (1997) 12. Borbely, J.S., George, M.C., Lombardi, L.D., Weel, M., Fitzakerley, D.W., Hessels, E.A.: . Phys. Rev. A 79, 060,503 (2009) 13. Brown, G.E., Langer, J.S., Schaeffer, G.W.: Lamb shift of a tightly bound electron. I. Method. Proc. R. Soc. London, Ser. A 251, 92–104 (1959) 14. Brown, G.E., Mayers, D.F.: Lamb shift of a tightly bound electron. II. Calculation for the K-electron in Hg. Proc. R. Soc. London, Ser. A 251, 105–109 (1959) 15. Cheng, T.K., Johnson, W.R., Sapirstein, J.: Screend Lamb-shift calculations for Lithiumlike Uranium, Sodiumlike Platimun, and Copperlike Gold. Phys. Rev. Lett. 23, 2960–63 (1991) 16. Clarke, J.J., van Wijngaaren, W.A.: . Phys. Rev. A 67, 12,506 (2003) 17. Curdt, W., Landi, E., Wilhelm, K., Feldman, U.: Wavelength measurements of heliumlike 1s2s 3 S1 1s2p 2 P0;2 transitions in Ne 8C , Na9C , M g 10C , and Si 12C emitted by solar flare plasmas. Phys. Rev. A 62, 022,502.1–7 (2000) 18. Desiderio, A.M., Johnson, W.R.: Lamb Shift and Binding Energies of K Electrons in Heavy Atoms. Phys. Rev. A 3, 1267–75 (1971) 19. DeVore, T.R., Crosby, D.N., Myers, E.G.: Improved Measurement of the 1s2s 1 S0 1s2p 3 P1 Interval in Heliumlike Silicon. Phys. Rev. Lett. 100, 243,001 (2008) 20. Drake, G.W.F.: Theoretical energies for the n D 1 and 2 states of the helium isoelectronic sequence up to Z D 100. Can. J. Phys. 66, 586–611 (1988) 21. George, M.C., Lombardi, L.D., Hessels, E.A.: Precision Microwave Measurement of the 2 3 P1 2 3 P0 Interval in Atomic Helium: A determination of the Fine-Structure Constant. Phys. Rev. Lett. 87, 173,002 (2001) 22. Giusfredi, G., Pastor, P.C., Natale, P.D., Mazzotti, D., deMauro, C., Fallani, L., Hagel, G., Krachmalnicoff, V., Ingusio, M.: Present status of the fine structure of the 2 3 P helium level. Can. J. Phys. 83, 301–10 (2005) 23. Grotsch, H.: . Phys. Rev. Lett. 24, 39 (1970) 24. Häffner, H., Beier, T., Hermanspahn, N., Kluge, H.J., Quint, W., Stahl, S., Verdú, J., Werth, G.: High-Acuuracy Measurements of the Magnetic Anomaly of the Electron Bound in Hydrogenlike Carbon. Phys. Rev. Lett. 85, 5308–11 (2000)

References

169

25. Indelicato, P., Gorciex, O., Desclaux, J.P.: Multiconfiguration Dirac-Fock studies of twoelectron ions: II. Radiative corrections and comparison with experiment. J. Phys. B 20, 651–63 (1987) 26. Jentschura, U.D., Mohr, P.J., Soff, G.: Calculation of the Electron Self-Energy for Low Nuclear Charge. Phys. Rev. Lett. 61, 53–56 (1999) 27. Jentschura, U.D., Mohr, P.J., Tan, J.N., Wundt, B.J.: Fundamental Constants and Tests of Theory in Rydberg States of Hydrogenlike Ions. Phys. Rev. Lett. 100, 160,404 (2008) 28. Kukla, K.W., Livingston, A.E., Suleiman, J., Berry, H.G., Dunford, R.W., Gemmel, D.S., Kantor, E.P., Cheng, S., Curtis, L.J.: Fine-structure energies for the 1s2s 3 S 1s2p 3 P transition in heliumlike ArC . Phys. Rev. A 51, 1905–17 (1995) ˚ en, B., Salomonson, S., M˚artensson-Pendrill, A.M.: QED procedure applied 29. Lindgren, I., As´ to the quasidegenerate fine-structure levels of He-like ions. Phys. Rev. A 64, 062,505 (2001) 30. Lindgren, I., Persson, H., Salomonson, S., Labzowsky, L.: Full QED calculations of twophoton exchange for heliumlike systems: Analysis in the Coulomb and Feynman gauges. Phys. Rev. A 51, 1167–1195 (1995) 31. Lindgren, I., Persson, H., Salomonson, S., Ynnerman, A.: Bound-state self-energy calculation using partial-wave renormalization. Phys. Rev. A 47, R4555–58 (1993) ˚ en, B.: The covariant-evolution-operator method in bound32. Lindgren, I., Salomonson, S., As´ state QED. Physics Reports 389, 161–261 (2004) 33. Marrs, R.E., Elliott, S.R., Stöhkler, T.: Measurement of two-electron contributions to the ground-state energy of heliumlike ions. Phys. Rev. A 52, 3577–85 (1995) 34. Mohr, P.: Self-energy correction to one-electron energy levels in a strong Coulomb field. Phys. Rev. A 46, 4421–24 (1992) 35. Mohr, P., Taylor, B.: CODATA recommended values of the fundamental constants 2002. Rev. Mod. Phys. 77, 1–107 (2005) 36. Mohr, P., Taylor, B.: CODATA recommended values of the fundamental constants 2002. Rev. Mod. Phys. (2008) 37. Mohr, P., Taylor, B.N.: CODATA recommended values of the fundamental physical constants: 1998. Rev. Mod. Phys. 72, 351–495 (2000) 38. Mohr, P.J.: Numerical Evaluation of the 1s1=2 -State Radiative Level Shift. Ann. Phys. (N.Y.) 88, 52–87 (1974) 39. Mohr, P.J.: Self-Energy Radiative Corrections. Ann. Phys. (N.Y.) 88, 26,52 (1974) 40. Mohr, P.J.: Lamb Shift in a Strong Coulomb Field. Phys. Rev. Lett. 34, 1050–52 (1975) 41. Mohr, P.J.: Self-energy of the n D 2 states in a strong Coulomb field. Phys. Rev. A 26, 2338– 54 (1982) 42. Mohr, P.J.: Quantum electrodynamics of high-Z few-electron atoms. Phys. Rev. A 32, 1949–57 (1985) 43. Mohr, P.J., Plunien, G., Soff, G.: QED corrections in heavy atoms. Physics Reports 293, 227–372 (1998) 44. Mohr, P.J., Sapirstein, J.: Evaluation of two-photon exchange graphs for excited states of highly charged heliumlike ions. Phys. Rev. A 62, 052,501.1–12 (2000) 45. Mohr, P.J., Soff, G.: Nuclear Size Correction to the Electron Self Energy. Phys. Rev. Lett. 70, 158–161 (1993) 46. Myers, E.G., Margolis, H.S., Thompson, J.K., Farmer, M.A., Silver, J.D., Tarbutt, M.R.: Precision Measurement of the 1s2p 3 P2 3 P1 Fine Structure Interval in Heliumlike Fluorine. Phys. Rev. Lett. 82, 4200–03 (1999) 47. Myers, E.G., Tarbutt, M.R.: Measurement of the 1s2p 3 P0 3 P1 fine structure interval in heliumlike magnesium. Phys. Rev. A 61, 010,501R (1999) 48. Persson, H., Labzowsky, I.L.L.N., Plunien, G., Beier, T., Soff, G.: Second-order self-energy– vacuum-polarization contributions to the Lamb shift in highly charged few-electron ions. Phys. Rev. A 54, 2805–13 (1996) 49. Persson, H., Lindgren, I., Salomonson, S., Sunnergren, P.: Accurate vacuum-polarization calculations. Phys. Rev. A 48, 2772–78 (1993) 50. Persson, H., Salomonson, S., Sunnergren, P., Lindgren, I.: Two-electron Lamb-Shift Calculations on Heliumlike Ions. Phys. Rev. Lett. 76, 204–07 (1996)

170


51. Persson, H., Salomonson, S., Sunnergren, P., Lindgren, I.: Radiative corrections to the electron g-factor in H-like ions. Phys. Rev. A 56, R2499–2502 (1997) 52. Plante, D.R., Johnson, W.R., Sapirstein, J.: Relativistic all-order many-body calculations of the n D 1 and n D 2 states of heliumlike ions. Phys. Rev. A 49, 3519–30 (1994) 53. Riis, E., Sinclair, A.G., Poulsen, O., Drake, G.W.F., Rowley, W.R.C., Levick, A.P.: Lamb shifts and hyperfine structure in 6 LiC and 7 LiC : Theory and experiment. Phys. Rev. A 49, 207–20 (1994) 54. Schweppe, J., Belkacem, A., Blumenfield, L., Clayton, N., Feinberg, B., Gould, H., Kostroun, V.E., Levy, L., Misawa, S., Mowat, J.R., Prior, M.H.: Measurement of the Lamb Shift in Lithiumlike Uranium (U C89 /. Phys. Rev. Lett. 66, 1434–37 (1991) 55. Sthöhlker, T., Beyer, H.F., Gumberidze, A., Kumar, A., Liesen, D., Reuschl, R., Spillmann, U., Trassinelli, M.: Ground state Lamb-shift of heavy hydrogen-like ions: status and prspectives. Hyperfine Interactions 172, 135–40 (2006) 56. Stöhlker, T., Mokler, P.H., et al., K.B.: Ground-state Lamb Shift for Hydrogenlike Uranium Measured at the ESR Storage Ring. Phys. Rev. Lett. 14, 2184–87 (1993) 57. Yerokhin, K.P.V.A.: Fine Structure of Heliumlike Ions and Determination of the Fine Structure Constant. Phys. Rev. Lett. 104, 70,403 (2010) 58. Yerokhin, V.A., Artemyev, A.N., Shabaev, V.M., Sysak, M.M., Zherebtsov, O.M., Soff, G.: Evaluation of the two-photon exchange graphs for the 2p1=2 2s transition in Li-like Ions. Phys. Rev. A 64, 032,109.1–15 (2001) 59. Yerokhin, V.A., Indelicato, P., Shabaev, V.M.: Self-energy Corrections to the Bound-Electron g Factor in H-like Ions. Phys. Rev. Lett. 89, 134,001 (2002) 60. Yerokhin, V.A., Indelicato, P., Shabaev, V.M.: Two-Loop Self-Energy Correction in High-Z Hydrogenlike Ions. Phys. Rev. Lett. 91, 073,001 (2003) 61. Yerokhin, V.A., Indelicato, P., Shabaev, V.M.: Evaluation of the self-energy correction to the g Factor of S states in H-like Ions. Phys. Rev. A 69, 052,203 (2004) 62. Zelevinsky, T., Farkas, D., Gabrielse, G.: Precision Measurement of the Three 23 PJ Helium Fine Structure Intervals. Phys. Rev. Lett. 95, 203,001 (2005)

Part III

Quantum-Electrodynamics Beyond Two-Photon Exchange: Field-Theoretical Approach to Many-Body Perturbation Theory

Chapter 8

Covariant Evolution Combined with Electron Correlation

In Part I, we have considered some standard methods for many-body calculations on atomic systems. These methods are well developed and can treat certain electroncorrelation effects to essentially all orders of perturbation theory. In Part II, we have considered three different methods for numerical QED calculations on bound systems, which have been successfully applied to various problems. All these methods are, however, in practice limited to one- and two-photon exchange, implying that electron correlation can only be treated in quite a restricted way. For many systems the electron correlation is of great importance, and to evaluate the QED effects accurately, it may be necessary to take into account also the combination of QED and correlational effects, which has not been done previously. In Part III, we demonstrate that one of the methods presented in the previous part, the covariant-evolution-operator method, forms a suitable basis for a combined QED–MBPT procedure.1 This leads to a perturbative procedure that is ultimately equivalent to an extension of the relativistically covariant Bethe–Salpeter equation, valid also in the multireference case and referred to as as the Bethe– Salpeter–Bloch equation. In this work, we normally use the Coulomb gauge, and we shall apply the equal-time approximation, discussed in Chap. 6. In Chap. 10, we illustrate how this procedure can be implemented and give some numerical results.

8.1 General Single-Photon Exchange In the treatment of single-photon exchange in Chap. 6, the incoming state was assumed to be unperturbed. We shall now generalize this treatment and allow the incoming state to be perturbed, involving particle as well as hole states. As mentioned, we shall deal particularly with the Coulomb gauge, where the total interaction is according to (4.57) separated into a Coulomb and a transverse part (see Fig. 8.1). I C D ICC C ITC :

1

(8.1)

The treatment in Part III is largely based on [3, 6, 7] and the thesis of Daniel Hedendahl [2].


173

174

8 Covariant Evolution Combined with Electron Correlation

Fig. 8.1 In the Coulomb gauge, the single-photon exchange is separated into a Coulomb and a transverse (Breit) part

x

x0

6

6

!3 r 6 1 !1 t 6

!4 6s 2 !2 6u

6

6

E

x0

x00

x

x0

6

6

!3 !4 6s r 6 z 1 2 !1 !2 6u t 6

6

E

6

x0

x00

The corresponding single-photon potential is similarly separated into: Vsp D VC C VT :

(8.2)

We start with the transverse part and consider the Coulomb part later.

8.1.1 Transverse Part The kernel of the transverse part of the single-photon exchange in Coulomb gauge according to (6.6) is given by: i SF .x; x1 / i SF .x 0 ; x2 / .i/ITC .x2 ; x1 / i SF .x1 ; x0 / i SF .x2 ; x00 / e.jt1 jCjt2 j/ : (8.3) The external time dependence is (with the notations in Figs. 8.1 and 8.2) in the equal-time approximation in analogy with the previous case (6.9): eit .!3 C!4 "r "s / eit0 .!1 C!2 "t "u / : As before, we can argue that in the limit ! 0 !1 C !2 D !3 C !4 D E, i.e., equal to the initial energy, and the dependence becomes: eit .E"r "s / eit0 .E"t "u / : We then have the relation: UT .t; t0 / D eit .EH0 / MT eit0 .EH0 / ;

(8.4)

where MT is the corresponding Feynman amplitude, defined as before (6.11). This yields: MT .x; x 0 I x 0 ; x 00 / D

ZZZZ Z 1 d!1 d!2 d!3 d!4 dz iSF .!3 I x; x 1 / 2 2 2 2 2 2 iSF .!4 I x 0 ; x 2 / .i/ITC .zI x 2 ; x 1 / iSF .!1 I x 1 ; x 0 / iSF .!2 I x 2 ; x 00 /2 .!1 z !3 / 2 .!2 C z !4 / (8.5)

8.1 General Single-Photon Exchange Fig. 8.2 Time-ordered evolution-operator diagrams for single-photon exchange, transverse part

175

x

r 6O ˙ O ˙6s

!4 6s !3 z r 6 2 x0

!2 6u 1 !1 t 6 t 6O ˙ O ˙6u

x

0

x

r 6O ˙ O ˙6s

r 6 1

-z

t 6 x00

x0

6s 2

6u

t 6O ˙ O ˙6u

E

x0

x00

E

leaving out the internal space integrations. (The factor of 1/2 is, as before, eliminated when a specific matrix element is considered.) After integrations over !2 ; !3 ; !4 , the amplitude becomes: MT .x; x

0

; x 0 ; x 00 /

ZZ 1 d!1 dz D iSF .!1 zI x; x 1 / iSF .E !1 zI x 0 ; x 2 / 2 2 2 .i/ITC .zI x 2 ; x 1 / iSF .!1 I x 1 ; x 0 / iSF .E !1 I x 2 ; x 00 /: (8.6)

Inserting the expressions for the electron propagator (4.10) and the interaction (4.46), a specific matrix element becomes: * ˇ Z Z ˇ dz 1 1 d!1 ˇ hrsjMTjabi D rs ˇi ˇ 2 2 !1 z "r C ir E !1 C z "s C is ˇ + Z 1 2c 2 d fTC ./ ˇˇ 1 ˇab ; !1 "t C it E !1 "u C iu z2 c 2 2 C i ˇ (8.7) where fTC is the transverse part of the f function in Coulomb gauge (4.60). Integration over z now yields – in analogy with the treatment in Chap. 6 MT D .i/2

Z c d fTC ./

times the propagator expressions 1 E " r "s

1 1 C !1 "r .c i /r E !1 "s .c i /s

and 1 E "t "u

1 1 C : !1 "t C it E !1 "u C iu

176


We have now four combinations that contribute depending on the sign of the orbital energies (after integration over !1 ): sgn."r / ¤ sgn."t / W

sgn."t / "t "r .c i /r

sgn."s / D sgn."t / W

sgn."t / E "t "s .c i /s

sgn."u / D sgn."r / W

sgn."u / E "r "u .c i /r

sgn."u / ¤ sgn."s / W

sgn."u / "u "s .c i /s

(8.8)

times .i/. The Feynman amplitude for the transverse part of the single-photon exchange now becomes: MT D .E/ iVT .E/ .E/;

(8.9)

where .E/ is the resolvent (2.64). This yields for the present process: ˇ ˛ ˝ ˇ rs ˇMT .E/ˇtu D

ˇ E D ˇ i 1 ˇ ˇ ; rs ˇ VT .E/ ˇtu E "r "s E "t "u

(8.10)

where VT .E/ is now the generalized transverse-photon potential: ˇZ ˇ ˛ ˇ ˝ ˇ C ˇ ˇ ˇ rs VT .E/ tu D rs ˇ c d fT ./ ˙

t˙ r "t "r ˙ c ˇ ˇ u˙ r˙ u ˙ s t˙ s˙ ˇ tu : ˙ ˙ ˙ E "t "s c E "r "u c "u "s ˙ c ˇ

(8.11)

Here, t˙ etc. represent projection operators for particle/hole states. The upper or lower sign should be used consistently in each term, inclusive of the sign in the front, but all combinations of upper and lower signs in the four term should be used, corresponding to the 16 time-ordered combinations, shown in Fig. 8.3. It should be noted that the expression above is valid also for the entire interaction in any covariant gauge, using the appropriate f function. We shall now illustrate the potential (8.11) by giving explicit expressions in a few cases.

8.1 General Single-Photon Exchange r t34 > t2 , the time integrations yield: .i/

3

Z

1

t34

id2 t2

dt2 e

Z

1 t2

id34 t34

dt34 e

Z

1

t2

dt1 eid1 t1 :

(8.21)

Together with the alternative time ordering 1 $ 2, this becomes: 1 d1234 d34

1 1 C d1 d2

(8.22)

180


with the notations of Appendix I, which is identical to the result (8.20). Note that this is NOT in agreement with the standard Goldstone rules of MBPT [4]. Single hole in and out (t,s) s 6

6 r

6

u

u

s

6

t

r

t E

E

The potential (8.11) yields: ˇZ ˇ ˛ ˇ ˝ ˇ 1 1 rs ˇVT .E/ˇtu D rs ˇˇ c d fTC ./ C "t "r c E "t "s C c

ˇ ˇ 1 1 ˇ tu : C C E "u "r c "u "s C c ˇ (8.23)

Using the notations of Appendix I: d1 D "t "r cI

d2 D "u "s C cI

d4 D "b "u C cI d134 D E "r "u cI

d3 D "a "t cI

d34 D E "t "u I

d234 D E "t "s CcI

d1234 D E "r "s ;

the bracket above becomes:

1 1 1 1 d1234 d34 C C D d1 d234 d2 d134 d134 d234

1 1 d2 d 1

(8.24)

and the denominators of the Feynman amplitude (8.10): 1 d134 d234

1 1 d2 d1

1 D d1 d2

which agrees with the rules of Appendix I.

1 d134

1 d234

;

(8.25)

8.2 General QED Potential

181 r; s < 0

No pairs r

s

t

u

r

t; u < 0 r

s

t

t; u; r; s < 0 s

u t

u

t ?

?u

r ?

?s

Fig. 8.4 Same as Fig. 8.3 for the Coulomb interaction

8.1.2 Coulomb Interaction The Coulomb part of the interaction is obtained in a similar way (see Fig. 8.4). In analogy with (8.5), we now have: ZZ Z 1 d!1 d!3 dz MC D iSF .!1 / iSF .E0 !1 / 2 2 2 2 .i/ICC iSF .!3 / iSF .E0 !3 / leaving out the space coordinates. After z integration, using (4.63b), and with the explicit form of the propagators this leads to: ˇ ZZ ˇ 1 d!1 d!3 1 hrsjMCjabi D rs ˇˇ i 2 2 !1 "r C ir E0 !1 "s C is ˇ ˇ 1 1 ˇ ab VC !3 "t C it E0 !3 "u C iu ˇ ˇ ˇ ˇ ˇ 1 i ˇ ab ; D rs ˇˇ˙ VC (8.26) E0 "r "s E0 "t "u ˇ where VC is the Coulomb interaction (2.109). Here, the plus sign is used if sgn."t / D sgn."u / D sgn."r / D sgn."s / and the minus sign if sgn."t / D sgn."u / ¤ sgn."r / D sgn."s /.

8.2 General QED Potential We shall now see how the potential above can be extended to include also crossing Coulomb interactions as well as various radiative effects.

8.2.1 Single Photon with Crossed Coulomb Interaction We start by considering a transverse photon with a crossing Coulomb interaction (Fig. 8.5), using the Coulomb gauge.

182


Fig. 8.5 Feynman diagram representing the exchange of a retarded covariant photon with crossing Coulomb interaction

6

x

6

r 4 !1 v 1

!3

t

!5

!4

? !6

6

E

2 w !2 3 u x00

6

6

x

s

z

x0

x0

6

!3

r

!4

x0 s 3 w !2 2

1 z6 !1 v !6 u 4 t !5 x0 x00

6

E

6

The Feynman amplitude becomes for the left diagram: ZZ ZZ ZZ Z d!1 d!2 d!3 d!4 d!5 d!6 dz 1 M.x; x 0 I x 0 ; x 00 / D 2 2 2 2 2 2 2 2 iSF .!3 I x; x 4 / iSF .!4 I x 0 ; x 2 / iSF .!1 I x 4 ; x 1 / iSF .!2 I x 2 ; x 3 / iSF .!5 I x 1 ; x 0 / iSF .!6 I x 3 ; x 00 / .i/ITC .zI x 4 ; x 3 / .i/VC .x 2 ; x 1 / 2 .!1 z !3 / 2 .!6 C z !2 / 2 .E !5 !6 / 2 .!1 C!4 !5 !2 /:

(8.27)

Integrations over !3 ; !4 ; !5 ; !6 lead in the adiabatic limit to !3 D !1 z;

!4 D E !1 C z;

and to M.x; x 0 I x 0 ; x 00 / D

1 2

ZZ

!5 D E !2 C z;

!6 D !2 z

d!1 dz iSF .!1 zI x; x 4 / iSF .E !1 C zI x 0 ; x 2 / 2 2

iSF .!1 I x 4 ; x 1 /iSF .!2 I x 2 ; x 3 /iSF .E !2 C zI x 1 ; x 0 / iSF .!2 zI x 2 ; x 00 / .i/ITC .zI x 2 ; x 3 / .i/VC .; x 4 ; x 3 /: (8.28) More explicitly the electron propagators become: 1 1 1 1 C .!1 "v C iv / .E "r "s / .!1 z "r C ir / .E !1 C z "s C is / 1 1 1 1 C : .!2 w C iw / .E "t "u / .E !2 C z "t C it / .!2 z "u C iu /

(8.29) The integrations over !1 ; !2 lead in analogy with (8.8) to v˙ r v˙ r sgn."v / ¤ sgn."r / W ˙ DW ˙ "v z "r i a v ˙ s˙ v ˙ s˙ DW ˙ sgn."v / D sgn."s / W ˙ E "v C z "s ˙ i b˙


183

sgn."w / D sgn."w / W sgn."w / ¤ sgn."u / W

w˙ t˙ w˙ t˙ DW ˙ E "w C z "t ˙ i c˙ w˙ u w˙ u ˙ DW ˙ "w z "u i d ˙

(8.30)

times .i/2 . Here, the first two terms should be combined with the last two, and the propagators (8.28) reduce to: 1 v˙ r v˙ s˙ 1 ˙ ˙ E "r "r E " t "u a b˙ w˙ t˙ w t w˙ u w u˙ : ˙ ˙ c˙ c d d˙

(8.31)

Upper or lower sign should be used consistently in all four operators in each product. We use here the notations: 8 ˆ ˆ A˙ D "v "r c < B˙ D E "v "s c ˆ C D E "w "t c ˆ : ˙ D˙ D "w "u c

8 a˙ D "v z "r ˙ i ˆ ˆ < b˙ D E "v C z "s ˙ i ˆ c D E "w C z "u ˙ i ˆ : ˙ d˙ D "w z "u ˙ i:

(8.32)

The photon interaction has one pole in each half-plane, and for the combinations where the electron propagator poles are in the same half-plane the z integration leads directly to the replacement a˙ ! A˙ etc. This part then becomes 1=.E "r "r / 1=.E "t "u / times

VC1

v˙ s˙ v˙ r D C A B˙

w˙ t˙ w˙ u : C C˙ D

(8.33)

Again, upper or lower sign should be used consistently in all four operators in each product. Expressing the Feynman amplitude in analogy with (8.10), ˇ ˛ ˝ ˇ rs ˇM.E/ˇtu D

ˇ E D ˇ i 1 ˇ ˇ ; rs ˇ VTC .E/ ˇtu E H0 E H0

(8.34)

the corresponding part of the potential becomes: Z hrsjVTC /1jtui D

c d fTC ./ hvsjVCjtwi hrwjVC1jvui :

(8.35)

When the electron propagators have one pole in each half-plane, we have to separate the propagators as before. For instance, the product

v˙ r w u v˙ r w t D a c "v z "r i E "w C z "t i

184


is rewritten as

v˙ r w t aCc

1 1 v˙ r w t 1 1 C C ) ; a c aCc A C

after z integration, or v˙ r w t E " w "t C "v "r

1 1 : C "v "r ˙ c E "w "t ˙ c

Similarly,

v˙ r w u˙ w u˙ v ˙ r D a d˙ "v z "r i "w z "u ˙ i

is rewritten as v˙ r w u˙ ad

1 1 v˙ r w u˙ 1 1 ) : a d˙ ad A D˙

But A D˙ D a d ; so this can also be written as:

v˙ r w u˙ : A D˙

All similar combinations lead to v˙ r w t 1 1 v˙ r w u˙ v˙ s˙ w t VC 2 D C aCc A C A D˙ B˙ C v˙ s˙ w u˙ 1 1 C ; (8.36) bCd B˙ D˙ where upper and lower sings are used consistently in all four operators in each term. The notations are defined in (8.32). This complete expression is quite complicated, particularly due to the denominator a C c. Equations (8.33) and (8.36) represent the complete potential for all 64 timeordered diagrams, corresponding to the Feynman diagram in Fig. 8.5: Z (8.37) hrsjVTCjtui D c d fTC ./ hvsjVCjtwi hrwjVC1 C VC 2jvui : We can simplify the results above by assuming that the incoming orbitals t; u are particle states. Then (8.33) reduces to vC sC wC tC v s w uC v r C vC r C C C (8.38) A BC CC AC B DC and (8.36) to v rC wC tC aCc

1 1 vC sC w uC 1 1 : C C AC CC bCd BC DC

(8.39)


185

With no pairs, we then have: vC sC wC tC vC sC wC tC D BC CC .E "v "s c/.E "w "t c/ with single pair (v) v rC wC tC aCc

1 1 C AC CC

v rC wC tC E " w " t C " v "r 1 1 C "v "r c E "w "t c

D

and double pairs (v; w) v rC w uC v rC w uC D : AC DC ."v "r c/."w "u c/ This corresponds to the time-ordered diagrams shown below, and the results are in agreement with the evaluation rules for time-ordered diagrams. Also here some diagrams are complicated to evaluate, due to the denominator a C c.

r

r v

s

s ?

t

w

t

v

r

w t

-

s

w

v -

u

u E

s

r

w u

E

t

6 u

v

E

E

Another, probably more reasonable approximation is to assume that the intermediate states v; w are particle states. Then only the simpler term (8.33) survives, yielding VC1C D

vC sC vC r C A BC

wC u wC tC C CC D

(8.40)

and the potential C jtui D hrsjVTC

Z c d fTC ./ hvsjVCjtwi hrwjVC1Cjvui :

(8.41)

This part of the potential can be generated by iterating the pair equation, as discussed in Chap. 10. This is true also for repeated Coulomb crossings.

186


8.2.2 Electron Self-Energy and Vertex Correction Next, we shall see how some radiative effects, previously treated in the S-matrix formalism (see Sect. 4.6), can be included in the QED potential. Since we are using the Coulomb gauge, we have to treat the Coulomb and transverse parts separately as in Sect. 4.6. We start with the transverse part, illustrated in Fig. 8.6 (left). The kernel is here (c.f. 4.84): (8.42) i SF .x; x2 / i SF .x2 ; x1 / .i/ITC .x2 ; x1 /; where ITC is the transverse part of the interaction (4.59). The Feynman amplitude becomes in analogy with previous cases (8.4) ZZ MSE .x/ D

d!1 d!2 dz iSF .!2 I x; x 2 / iSF .!1 I x 2 ; x 1 / .i/ITC .zI x 2 ; x 1 / 2 2 2 2 ."a !1 z/ 2 .!1 !2 C z/ (8.43)

integrated over the internal space coordinates and with the energy parameters given in the figure. After integration over the omegas, this becomes: Z MSE .x/

dz iSF ."a I x; x 2 / iSF ."a zI x 2 ; x 1 / .i/ITC .zI x 2 ; x 1 /: 2

(8.44)

The matrix element of the evolution operator is: hrjUSE .t/jai D eit ."a "r / hrtjMSE .x/jtai ;

(8.45)

which we can express ˇ ˛ ˇ ˛ ˝ ˇ eit ."a "r / ˝ ˇˇ r i˙."a /ˇa ; r ˇUSE .t/ˇa D " a "r

(8.46)

where ˙."a / is the self-energy operator (4.85) ˇ ˇZ ˇ dz ˇ iSF ."a zI x 2 ; x 1 / ITC .zI x 2 ; x 1 /ˇˇ ta : hrj˙."a /jai D rt ˇˇ 2 x Fig. 8.6 Diagram representing the transverse and Coulomb parts of the first-order self-energy of a bound electron in the covariant-evolution-operator formalism (c.f. Fig. 4.9)

r 2

!2

t

!1

1 a

(8.47)

x r z

2

t 1 a


187

Inserting the explicit expressions for the propagators, yields hrj˙."a /jaiTrans

ˇZ ˇ D rt ˇˇ

ˇ ˇ c d fTC ./ ˇ ta "a "t .c i/t ˇ

(8.48)

consistent with the diagonal S-matrix (4.89). It should be noted that the self-energy is diagonal in energy in the S-matrix formulation, due to the energy conservation, while also nondiagonal parts will appear in the covariant-evolution formulation. As we shall see, only the diagonal part is divergent and has to be renormalized, as will be discussed in Chap. 12. The Coulomb part of the self-energy (Fig. 8.6 right) becomes in analogy with the S-matrix result (4.95) hrj˙."a /jaiCoul

ˇ ˇ Z ˇ 1 e2 2 d sin r12 ˇˇ ˇ D rt ˇ 2 ˇ ta 2 4 0 r12 2 D

1 sgn."t / hrt jVC j tai 2

(8.49)

with summation over positive- as well as negative-energy states.

8.2.2.1 General Two-Electron Self-Energy We consider now a general self-energy operator (transverse part) in analogy with the general single-photon exchange in Sect. 8.1, illustrated in Fig. 8.7. (The Coulomb part can be treated similarly.) The kernel is now i SF .x; x2 / i SF .x2 ; x1 / .i/ITC .x2 ; x1 / i SF .x1 ; x0 / i SF .x 0 ; x00 / e.jt1 jCjt2 j/ (8.50)

x0

x !3 r 6 2 v 6 Fig. 8.7 General two-electron self-energy with incoming and outgoing electron propagators

x0

z

1 !1 t 6 E

!2 6u x00

188


and the Feynman amplitude MSE .x; x

0

I x 0 ; x 00 /

ZZZZ D

d!1 d!2 d!3 d!4 2 2 2 2

Z

dz iSF .!3 I x; x 2 / 2

iSF .!4 I x 2 ; x 1 / iSF .!1 I x 1 ; x 0 / iSF .!2 I x 0 ; x 00 / .i/ITC .zI x 2 ; x 1 / 2 .!1 z !4 / 2 .!4 C z !3 / 2 .E !1 !2 /:

(8.51)

After integrations over !2 ; !3 ; !4 , this becomes MSE .x; x

0

; x 0 ; x 00 /

ZZ D

d!1 dz iSF .!1 I x; x 2 / iSF .!1 zI x 2 ; x 1 / 2 2 .i/ITC .zI x 2 ; x 1 /iSF .!1 I x 1 ; x 0 /iSF .E !1 I x 0 ; x 00 / (8.52)

as before, leaving out the internal integrations. Integration over z now yields Z 2 c d fTC ./ MSE D .i/ times the propagator expressions 1 1 1 "r "v .i /v !1 "r C ir !1 "v .c i /v and 1 E "t "u

1 1 C : !1 "t C it E !1 "u C iu

(8.53)

The integration over !1 yields another factor of i, and this leads in analogy with (8.10) to: ˇ ˛ ˝ ˇ ruˇMSE .E/ˇtu D

i 1 ; hru jVSE .E/ j tui "r "v .c i /v E " t "u

(8.54)

where VSE .E/ is the potential ˇZ ˇ ˛ ˇ ˝ ˇ t˙ r r˙ u ˙ ˇ ˇ ˙ rs VSE .E/ tu D rs ˇˇ c d fTC ./ ˙ "t "r E "r "u ˇ ˇ t˙ v v˙ u˙ ˇ tu : "t "v ˙ c E "u "v c ˇ

(8.55)


189

If all states are particle states, we find that the bracket above becomes: 1 1 "r "v c D E "r "u E "u "v c .E "r "u /.E "u "v c/ and the Feynman amplitude 1 i ; .E "r "u /.E "u "v c/ E "t "u

(8.56)

in agreement with the evaluation rules for time-ordered diagrams, derived in Appendix I. Next, we consider some specific cases with virtual holes, and, as before, we apply the potential to a Coulomb interaction. Hole out (r) (remaining ones particle states) r v

r v

6 6u

t 6

6u

t 6

E

E

The Feynman amplitude (8.54) becomes: i "r "v c

1 1 " t "r E "u "v ck

1 : E " t "u

This corresponds to the two time-ordered diagrams in the marginal. Hole in (t) r v t

6u

6u

r 6

6 E

t

v 6 E

(8.57)

190


The Feynman amplitude becomes: i 1 1 1 C C "r "v c "t "r E "r " u "t "v c 1 1 ; E "u "v ck E "t "u

(8.58)

which can also be expressed as i

1 1 1 : C ."r "t /."t "v c/ .E "r "u /.E "u "v c/ E " t "u (8.59)

After some additional algebra, this can be shown to be identical to i 1 1 C ."t "v c/.E "r "u / E "u "v c "t "r

(8.60)

corresponding to the time-ordered diagrams in the marginal.

8.2.2.2 General Vertex Correction The general vertex correction (transverse part), illustrated in Fig. 8.8, leads to MVx .x; x 0 ; x 0 ; x 00 / D

ZZ

d!1 d!1 2 2

Z

dz iSF .!3 I x; x 2 / iSF .!3 zI x 2 ; x 3 / 2

iSF .E !3 I x 0 ; x 4 /.i/ITC .zI x 2 ; x 1 / .i/VC .x 4 ; x 3 / iSF .!1 zI x 3 ; x 1 / iSF .!1 I x 1 ; x 0 / iSF .E !1 I x 4 ; x 00 /:

(8.61)

x0

x

Fig. 8.8 General vertex correction with incoming and outgoing electron propagators

x0

r 2 w 3 v 1 t

6!3

!4 6s

6

4

6

!2 6u

6!1 E

x00


191

More explicitly, the electron propagators become: 1 1 1 1 C E "r "s !3 "w z iw !3 "r C ir E !3 "s C is times 1 1 1 1 : C E "t "u !1 "v z iv !1 "t C it E !1 "u C iu

(8.62)

If the energies of the orbitals v and w have the same sign, then the integration over z leads to: Z MVx D i

c d fTC ./

times 1 1 1 1 C E "r "s !3 "w .c i /w !3 "r C ir E !3 "s C is and 1 1 1 1 C (8.63) E "t "u !1 "v .c i /v !1 "t C it E !1 "u C iu and to the Feynman amplitude, in analogy with (8.54): ˇ ˛ ˝ ˇ rs ˇMVx .E/ˇtu D

ˇ E D ˇ i 1 ˇ ˇ ; rs ˇ VVx .E/ ˇtu E "r " s E "t "u

(8.64)

where VVx .E/ is the potential ˇ Z ˇ ˛ ˇ ˇ ˛˝ ˇ ˝ ˇ ˇ ˇ ˇ ˇ ˇ rs VVx .E/ tu D ws ˇVC vu rv c d fTC ./ s˙ w˙ r˙ w ˙ ˙ "r "w ˙ ck E "s "w c ˇ ˇ v˙ u˙ t˙ v ˇ wt : ˙ ˙ "t "v ˙ c E "u "v c ˇ

(8.65)

If the orbitals v and w are of different kind (particle or hole), the evaluation becomes more complicated. This case is expected to be less important. If all states are particle states, we find that the brackets above become: 1 1 ; E "s "w c E "u "v c

192


in agreement with the evaluation rules for time-ordered diagrams. If r is a hole state and the others particle states, we have instead

1 1 1 C "r "w c E "s "w c E "u "v c

D

1 E "r "s : ."r "w c/.E "s "w c/ E "u "v c

This leads to the denominators of the Feynman amplitude (8.64)

1 ."r "w c/.E "s "w c/.E "u "v c/.E "t "u /

and corresponds to the time-ordered diagram. r w6

6s 6u

v 6 t 6 E

8.2.3 Vertex Correction with Further Coulomb Iterations The Coulomb interactions of the vertex correction can be iterated before the photon interaction is closed, in the same way as for the retarded photon with crossed Coulomb, treated above, leading to diagrams of the type shown in Fig. 8.9. Assuming that the intermediate states v; w, as well as the states between the Coulomb interactions are particle states, the corresponding analytical expression is obtained from (8.64) by replacing the matrix element hwsjVCjvui by: hwsjVC

jxyi hxyj VCjvui : E "x "y c

r 6 2 w 6

Fig. 8.9 Vertex correction with double Coulomb interactions

6s 6y

x 6 v 6 1 t 6

6u E

8.3 Unification of the MBPT and QED Procedures

6

V QED

6

6 6

6 Vsp D

6

6

6

6

6 C

6

193

6

6

6 6

6 C

6 6

6 6C

6 6 6

6 6

Fig. 8.10 Feynman diagram representing the “QED potential,” V QED , in (8.66). The first diagram on the r.h.s. includes the Coulomb potential

8.2.4 General Two-Body Potential We can now form a general “two-body QED potential” by adding the contributions derived above: V QED D Vsp C VTC C VSE C VVx ;

(8.66)

where Vsp , as before, represents the combined Coulomb and transverse-photon exchange (8.2). This is illustrated in Fig. 8.10. Here, particle as well as holes are allowed in and out.

8.3 Unification of the MBPT and QED Procedures: Connection to Bethe–Salpeter Equation We shall now see how the general QED potentials, derived above by means of the field-theoretical Green’s operator, can be combined with the standard MBPT procedure, leading to a unified MBPT–QED procedure. The procedure is valid for an arbitrary (quasi-degenerate) model space and equivalent to an extension of the standard Bethe–Salpeter equation, referred to as the Bethe–Salpeter–Bloch equation, briefly mentioned in Sect. 6.9 and further discussed in the next chapter (see also [5]). The procedure is also applicable to systems with more than two electrons, as will be briefly discussed at the end of this chapter.

8.3.1 MBPT–QED Procedure The general potentials derived above – with possible hole states on the in- and outgoing lines – cannot be used iteratively in the way discussed in Chap. 6. Therefore, it cannot be used directly in a Bloch equation, like that in (6.106). For that purpose, we shall insert one extra Coulomb interaction, when holes are present, leading to the replacements V QED ) V QED C V QED Q .E/VC C VC Q .E/V QED C VC Q .E/V QED Q .E/VC (8.67)

194


6

6

6

6)

6

6

V QED

6

V QED

6

6

6

+ 6

6

V QED

6

6

6

6VP 6 + 6VP 6 + VC

6

6

6

6

6

6VP 6 6VP 6 6

6

Fig. 8.11 Illustration of the modified potential (8.67), which can be iterated. It has only positiveenergy states in and out and is free from the Brown–Ravenhall effect

6

˝ QED

6

6

D

6

6

6C

6

6

6Q

6

6

6

V QED

C

˝ QED

6 6 6

6

P

6

ı ˝ QED ıE QED

Veff

6

Fig. 8.12 Graphical representation of the single-photon Bloch equation (8.68). The last diagram represents the “folded” term, i.e., the last term of the equation. This equation can be compared with the Bethe–Salpeter equation in Fig. 9.4, valid only in the single-reference case, where there is no folded contribution. The order-by-order expansion of this equation is illustrated in Fig. 8.13

illustrated in Fig. 8.11. This potential has only particle states (positive energy) in and out, and can therefore be used iteratively in a Bloch equation. When we have a single negative-energy state in the output, we can have a vanishing denominator of the final resolvent, which leads to a singularity of the Brown–Ravenhall type [1]. As demonstrated, though, at the beginning of this chapter, such singularities cancel when combined with the general potential. But then it is of vital importance that the potential and resolvent appear in “matching pairs.” This will always be the case when the modified potential (8.67) is applied. Inserting the modified potential (8.67) into the Bloch equation (6.106) leads to

˝ QED D 1 C Q V QED ˝ QED C

ı ˝ QED QED Veff ; ıE

(8.68)

QED where Veff D P V QED .Heff /˝ QED P is the corresponding effective interaction (6.139). In the last folded term only the last interaction, with the corresponding resolvent, is differentiated [see (6.105)]. The modified potential (8.67) is here regarded as a single unit. This equation is illustrated graphically by the Dyson type of equation in Fig. 8.12. The iterative expansion of the equation is displayed in Fig. 8.13. Solving the equation iteratively is equivalent to solving the corresponding version of the Bethe–Salpeter–Bloch equation [see (6.140) and (9.30)]. The potential discussed above represents the dominating part of the QED effects. In order to get further, also irreducible combinations of transverse interactions

8.3 Unification of the MBPT and QED Procedures

6 ˝ QED 6

D

6

6C

6

6

195

6 V QED 6 6

C

6

6

6

6

6C

6

6

+ folded

Fig. 8.13 Graphical representation of the order-by-order expansion of the Bloch equation in Fig. 8.12

C6

6 6 6 6+ 6

6

6 6

˝ QED 6 6

6˝I 6

6

6

D

6

6

6 V QED

6 6 6 6 6 6+ 6 6 6

6 6 6 6 +..+ 6 P 6 6 6

6 6 6

Fig. 8.14 Graphical representation of the Bloch equation (8.70), where a standard pair function (˝I ) is combined with a QED potential

should be included (see Fig. 6.6). Formally, we can express the corresponding Bloch equation: ˝ QED D 1 C Q V QED ˝ QED C

ı ˝ QED QED Veff ; ıE

(8.69)

where V QED is the QED potential, based on the generalized multiphoton potenQED tial, used previously (Fig. 6.6), and Veff is the corresponding effective interaction (6.139). This corresponds to the full Bethe–Salpeter–Bloch equation (without singles). For the time being, though, it does not seem feasible to go beyond a single transverse photon. However, the two-photon exchange can be approximated by including one retarded and one instantaneous transverse (Breit) interaction. The potential (8.67) can also be combined with standard pair functions without virtual pairs (Fig. 2.6). This leads to the Bloch equation ˝ QED D ˝I C Q V QED ˝ QED C

ı ˝ QED QED Veff ıE

(8.70)

illustrated in Fig. 8.14 (and analogously in the generalized case). This implies that the Coulomb interaction is iterated to much higher order than the transverse interaction. But since the Coulomb interaction normally dominated heavily over the transverse interaction, this procedure usually represents a much faster way of generating a perturbative scheme than that represented by (8.68) and Fig. 8.12. In the next section, we shall describe how the QED potential (8.67) can be used in a coupled-cluster expansion, in analogy with the standard procedure of MBPT, described in Sect. 2.5. Then also single-particle effects can be included in a systematic way, and the procedure would, in principle, be fully equivalent to the complete

196


Bethe–Salpeter–Bloch equation with singles, applicable also to open-shell systems. This approach will also make it possible to apply the procedure to more than two electrons. In the next chapter, we analyze the Bethe–Salpeter and the Bethe–Salpeter–Bloch equations further. In Chap. 10, we discuss how the iterative procedure discussed here with a single transverse photon can be implemented and give some in numerical illustrations. The renormalization procedure is discussed in Chap. 12.

8.4 Coupled-Cluster-QED Expansion With the interactions derived above, we can construct an effective QED-CoupledCluster procedure in analogy with that used in standard MBPT, described in Sect. 2.5 (see [7]). Considering the singles-and-doubles approximation (2.105) S D S1 C S 2 ;

(8.71)

the MBPT/CC equations are illustrated in Fig. 2.8. In order to obtain the corresponding equations with the covariant potential (6.6), we make the replacements illustrated in Fig. 8.15, which leads to the equations illustrated in Fig. 8.16. The CC-QED procedure can also be applied to systems with more than two electrons. For instance, if we consider the simple approximation (2.101) 1 ˝ D 1 C S2 C fS22 g; 2 then we will have in addition to the pair function also the coupled-cluster term, illustrated in Fig. 8.17 (left). Here, one or both of the pair functions can be replaced by the QED pair function in Fig. 8.12 (right) to insert QED effects on this level. In addition, of course, single-particle clusters can be included, as in the two-particle case discussed above (Fig. 8.15).

6

)

6

6

x

=

6

6 6

6

6

6

6

)

+

6 6

6 + 6 ?

+

6

6

6

6

6

6

6

6

6

6 6 6 6

)

+

6 6

+

6

6

6

6

6

Fig. 8.15 Replacements to be made in the CC equations in Fig. 2.8 to generate the corresponding CC-QED equations (c.f. Figs. 6.4 and 6.5). The wavy line in the second row represents the modified potential (8.67) with only particle states in and out

8.4 Coupled-Cluster-QED Expansion

6

S1 W

197

6

6

=

6

6

+

6

=

6

6

6

6

6

+

+

6

6

+ +

6+

6

6

6

6

6

6

6 6 6

6

6

6

6

6

6

+

6 6

W1

6

6

6

6

x

6

6

6

6

S2 W

+

6

6

6

6

6

6

6

6

+

x

x

+

6

6

6

6

6

+

6

W2

6

6

6

6

6

6+

6

6

6

6

6

6

+

6 6

+

x

6

6

6

6

+ +

6

6

6

6

W1

+

W2

Fig. 8.16 Diagrammatic representation of the QED-coupled-cluster equations for the operators S1 and S2 . The second diagram in the second row and the diagrams in the fourth row are examples of coupled-cluster diagrams. The last diagram in the second row and the three diagrams in the last row represent folded terms (c.f. the corresponding standard CC equations in Fig. 2.8)

6

6 6

6

6

6 6

6

6

6 6

6

6

6 6

6

6

6 6

6

6

6 6

6

Fig. 8.17 Diagrammatic representation of the QED-coupled-cluster term 12 S22 with standard pair functions (left) and one and two inserted QED pair function, defined in Fig. 8.12, (right)

198


We can summarize the results obtained here in the following way: When all one- and two-particle effects are included, the MBPT–QED proce-

dure is fully compatible with the two-particle Bethe–Salpeter(–Bloch) equation – including singles. The advantage of the MBPT–QED procedure is – thanks to the complete compatibility with the standard MBPT procedure – that the QED potentials need to be included only in cases where the effect is expected to be sufficiently important. The procedure described here is based on the use of the Coulomb gauge (6.41), and therefore not strictly covariant. As mentioned, however, in practice, it is equivalent to a fully covariant procedure, and, furthermore, it seems to be the only feasible way for the time being to treat effects beyond two-photon exchange in a systematic fashion.

References 1. Brown, G.E., Ravenhall, D.G.: On the Interaction of Two electrons. Proc. R. Soc. London, Ser. A 208, 552–59 (1951) 2. Hedendahl, D.: Towards a Relativistic Covariant Many-Body Perturbation Theory. Ph.D. thesis, University of Gothenburg, Gothenburg, Sweden (2010) 3. Hedendahl, D., Salomonson, S., Lindgren, I.: . Phys. Rev. p. (to be published) (2011) 4. Lindgren, I., Morrison, J.: Atomic Many-Body Theory. Second edition, Springer-Verlag, Berlin (1986, reprinted 2009) 5. Lindgren, I., Salomonson, S., Hedendahl, D.: Many-body-QED perturbation theory: Connection to the two-electron Bethe-Salpeter equation. Einstein centennial review paper. Can. J. Phys. 83, 183–218 (2005) 6. Lindgren, I., Salomonson, S., Hedendahl, D.: Many-body procedure for energy-dependent perturbation: Merging many-body perturbation theory with QED. Phys. Rev. A 73, 062,502 (2006) 7. Lindgren, I., Salomonson, S., Hedendahl, D.: Coupled clusters and quantum electrodynamics. ˇ arsky, J. Paldus, J. Pittner (eds.) Recent Progress in Coupled Cluster Methods: Theory In: P. C´ and Applications, pp. 357–374. Springer Verlag (2010)

Chapter 9

The Bethe–Salpeter Equation

In this chapter, we discuss the Bethe–Salpeter equation and its relation to the procedure we have developed so far. We start by summarizing the original derivations of the equation by Bethe and Salpeter and by Gell-Mann and Low, which represented the first rigorous covariant treatments of the bound-state problem. We demonstrate that this field-theoretical treatment is completely compatible with the presentation made here. The treatments of Bethe and Salpeter and of Gell-Mann and Low concern the single-reference situation, while our procedure is more general. We, later in this chapter, extend the Bethe–Salpeter equation to the multireference case, which will lead to what we refer to as the Bethe–Salpeter–Bloch equation in analogy with corresponding equation in MBPT.

9.1 The Original Derivations by the Bethe–Salpeter Equation The original derivations of the Bethe–Salpeter equation by Salpeter and Bethe [15] and by Gell-Mann and Low [9] were based on procedures developed in late 1940s for the relativistic treatment of the scattering of two or more particles by Feynman [7, 8], Schwinger [18, 19], Tomanaga [23] and others, and we here summarize their derivations.

9.1.1 Derivation by Salpeter and Bethe Salpeter and Bethe [15] start their derivation from the Feynman formalism of the scattering problem [7, 8], illustrated in terms of Feynman graphs. A Feynman diagram represents in Feynman’s terminology the “amplitude function” or “kernel” for the scattering process, which in the case of two-particle scattering, denoted K.3; 4I 1; 2/, is the probability amplitude for one particle propagating from one space-time point x1 to another x3 and the other particle from space-time x2 to x4 . For the process involving one irreducible graph G .n/ , i.e., a graph that cannot be I. Lindgren, Relativistic Many-Body Theory: A New Field-Theoretical Approach, Springer Series on Atomic, Optical, and Plasma Physics 63, c Springer Science+Business Media, LLC 2011 DOI 10.1007/978-1-4419-8309-1 9,

199

200

9 The Bethe–Salpeter Equation

3 3

G

.n/

4

5

6

7

8 6 2

1

G

.n/

5 7

4 6 8

G .m/ 1

2

Fig. 9.1 Examples of Feynman graphs representing scattering amplitudes in Eqs. (9.1) and (9.2) of the Salpeter–Bethe paper [15]. The first diagram is irreducible, while the second is reducible, since it can be separated into two allowed diagrams by a horizontal cut

3

4

=

K.3; 4I 1; 2/ 1

+

+

G

+

2

Fig. 9.2 Graphical representation of the expansion of the Feynman kernel in terms of irreducible graphs

separated into two simpler graphs, as illustrated in Fig. 9.1 (left part), the kernel is given by (in Feynman’s notations) ZZZZ .n/ K .3; 4I 1; 2/ D i d5 d8 KCa .3; 5/KCb .4; 6/ G .n/ .5; 6I 7; 8/ KCa .7; 1/KCb .8; 2/;

(9.1)

where KCa ; KCb represent free-particle propagators (positive-energy part). For a process involving two irreducible graphs, the kernel illustrated in the right part of the figure becomes: ZZZZ K .n;m/ .3; 4I 1; 2/ D i d5 d8 KCa .3; 5/KCb .4; 6/ G .n/ .5; 6I 7; 8/ K .m/.7; 8I 1; 2/:

(9.2)

This leads to the sequence illustrated in Fig. 9.2, where G represents the sum of all irreducible two-particle self-energy graphs. From this, Salpeter and Bethe arrived at an integral equation for the total kernel ZZZZ d5 d8 KCa .3; 5/KCb .4; 6/ K.3; 4I 1; 2/ D KCa .3; 1/KCb .4; 2/ i G .5; 6I 7; 8/ K.7; 8I 1; 2/:

(9.3)

9.1 The Original Derivations by the Bethe–Salpeter Equation x0

x

x

x0

=

G.x1 ; x10 I x0 ; x00 / x00

x0

+ x0

x00

201 x

x0

x2 x1

x20 ˙ .x2 ; x20 I x1 ; x10 / x10 G.x1 ; x1 I x0 ; x00 /

x0

x00

Fig. 9.3 Graphical representation of the integral equation (9.3) for the Feynman kernel of Salpeter and Bethe – identical to the Dyson equation for the two-particle Green’s function (Fig. 5.8)

This is the equation for the two-particle Greens function (5.80) in the form of a Dyson equation, in our notations written as: ZZZZ d4 x1 d4 x2 d4 x10 d4 x20 G.x; x 0 I x0 ; x00 / D G0 .x; x 0 I x0 ; x00 / C G0 .x; x 0 I x2 ; x20 / .i/˙ .x2 ; x20 I x1 ; x10 / G.x1 ; x10 I x0 ; x00 / (9.4) and depicted in Fig. 9.3 (see also Fig. 5.8). Note that the two-particle kernel K in the terminology of Feynman and Salpeter–Bethe corresponds to our Green’s function G, and the irreducible interaction G corresponds to our proper self-energy ˙ . The proper (or irreducible) self-energy is identical to the irreducible two-particle potential in Fig. 6.6. Furthermore, the electron propagators are in the Feynman– Salpeter–Bethe treatment free-particle propagators. Note that the intermediate lines in Fig. 9.3 represent a Green’s function, where the singularities are eliminated. Salpeter and Bethe then argued that a similar equation could be set up for the bound-state wave function. Since the free lines of the diagrams in the Feynman formulation represent free particles, they concluded that the first (inhomogeneous) term on the right-hand side could not contribute, as the bound-state wave function cannot contain any free-particle component. This leads in their notations to the homogeneous equation: ZZZZ .3; 4/ D i

d5 d8 KCa .3; 5/KCb .4; 6/ G .5; 6I 7; 8/ .7; 8/: (9.5)

This is the famous Bethe–Salpeter equation. In the Furry picture we use here, where the basis single-electron states are generated in an external (nuclear) potential, the inhomogeneous term does survive, and the equation becomes in our notations ZZZZ 0 0 d4 x1 d4 x2 d4 x10 d4 x20 .x; x / D ˚.x; x / C G0 .x; x 0 I x2 ; x20 / .i/˙ .x2 ; x20 I x1 ; x10 / .x1 ; x10 /:

(9.6)

202


x0

x

x0

x

=

+

x

x0

x2 x1

x20 ˙ .x2 ; x20 I x1 ; x10 / x10

˚

Fig. 9.4 Graphical representation of the inhomogeneous Bethe–Salpeter equation (9.6). ˙ represents the proper self-energy, which contains all irreducible interaction graphs and is identical to the irreducible two-particle potential in Fig. 6.6. This equation can be compared with that represented in Fig. 8.12, valid also in the multireference case

This is the inhomogeneous Bethe–Salpeter equation we shall use, and it is graphically depicted in Fig. 9.4.

9.1.2 Derivation by Gell-Mann and Low The derivation of Gell-Mann and Low [9] starts from the “Feynman two-body kernel,” used in the definition of the Green’s function (5.20) (in their slightly modified notations): ˇ ˛ ˝ ˇ K.x1 ; x2 I x3 ; x4 / D 0 ˇT Œ O H .x1 / O H .x2 / O H .x4 / O H .x3 /ˇ0 :

(9.7)

T is the time-ordering operator (2.27) and O H ; O H are the particle-field operators in the Heisenberg representation. 0 is the vacuum (ground state) of the interacting system in the Heisenberg picture,j0H i. In an Appendix of the same paper, Gell-Mann and Low derive a relation between the interacting (0 ) and the noninteracting (˚0 ) vacuum states (both in the interaction picture): c0 D

U.0; 1/˚0 ; h˚0 jU.0; 1/j˚0 i

(9.8)

which is the famous Gell-Mann–Low theorem (3.29), discussed previously. Here, c is a normalization constant (equal to unity in the intermediate normalization that we use). This can be eliminated by considering 1 D h0j0 i D

c2

h˚0 jU.1; 1/j˚0 i : h˚0 jU.1; 0/j˚0 i h˚0 jU.0; 1/j˚0 i

(9.9)

9.1 The Original Derivations by the Bethe–Salpeter Equation

203

Inserting the expression (9.8) into the kernel (9.7), using the relation (9.9), yields: K.x1; x2 I x3 ; x4/ D

h˚0 jU .1; 0/T Œ O H .x1 / O H .x2 / O H .x4 / O H .x3 /U.0; 1/j˚0 i ; h˚0 jU.1; 1/j˚0 i (9.10)

which is equivalent to the field-theoretical definition of the Green’s function G.x1 ; x2 I x3 ; x4 / in (5.20). Gell-Mann and Low then conclude that expanding the expression above in a perturbation series leads to the two-body kernel of Feynman in terms of Feynman diagrams, as we have performed in Chap. 5. This is identical to the expansion given by Salpeter and Bethe, and hence leads also to the integral equation (9.3). GellMann and Low then use the same arguments as Salpeter and Bethe to set up the Bethe–Salpeter equation (9.5) for the wave function. In addition, they argue that single-particle self-energy parts can easily be included by modifying the singleparticle propagators. The derivation of Gell-Mann and Low, which starts from the field-theoretical definition of the Green’s function, has a firm field-theoretical basis. This is true, in principle, also of the derivation of Salpeter and Bethe, which is based on Feynman diagrams for scattering of field-theoretical origin. In Sect. 9.1.3, we shall see how the Bethe–Salpeter equation can be motivated from the graphical form of the Dyson equation in Fig. 9.3.

9.1.3 Analysis of the Derivations of the Bethe–Salpeter Equation We can understand the Bethe–Salpeter equation graphically, if we let the Dyson equation in Fig. 9.3 act on the zeroth-order state, ˚.x0 ; x00 /, which we represent by two vertical lines without interaction. (The treatment can easily be extended to the situation, where the model function is a linear combination of straight products.) From the relation (6.7), we see that the electron propagator acting on an electron-field operator (with space integration) shifts the coordinates of the operator. Therefore, acting with the zeroth-order Green’s function on the model function shifts the coordinates of the function according to: ZZ ˚.x; x 0 / D d3 x 0 d3 x 00 G0 .x; x 0 I x0 ; x00 / ˚.x0 ; x00 /: (9.11) This is illustrated in Fig. 9.5 (left) and corresponds to the first diagram on the righthand side of Fig. 9.4. Similarly, operating with the full Green’s function in Fig. 9.5 on the model function leads to ZZ d3 x 0 d3 x 00 G.x; x 0 I x0 ; x00 / ˚.x0 ; x00 / (9.12) .x; x 0 / D illustrated in Fig. 9.5 (right). Then the entire equation (9.6), illustrated in Fig. 9.4, is reproduced.

204

9 The Bethe–Salpeter Equation x0

x x

x

0

G0 x0

x00

x0

x00

x

x

0

)

˚

x0 ˚

x0

x G

x00 x00

x0 ˚

)

Fig. 9.5 Graphical illustration of Eqs. (9.11) and (9.12)

The equation (9.12) is consistent with the definition of the classical Green’s function (5.1), which propagates a wave function from one space-time point to another – in our case from one pair of space-time point to another. This equation can also be expressed as an operator equation ˛ ˛ j .t; t 0 / D G.t; t 0 I t0 ; t00 / j .t0 ; t00 / ;

(9.13)

where G is the Green’s operator, introduced in Sect. 6.6. The coordinate representation of this equation ˛ ˝ ˛ ˛ hx; x 0j .t; t 0 / D x; x 0 jG.t; t 0 I t0 ; t00 /jx 0 ; x 00 hx 0 ; x 00j .t0 ; t00 /

(9.14)

is identical to (9.12). This implies that: The Green’s function is the coordinate representation of the Green’s operator. The four-times Green’s operator represents the time propagation of the two-

particle Bethe–Salpeter state vector. In the equal-time approximation, this is consistent with our previous result (6.50) and with our conjecture (6.29). It is of interest to compare the Bethe–Salpeter equation (9.6), depicted in Fig. 9.4, with the Dyson equation for the combined QED-electron correlation effects in Fig. 8.12. If in the latter more and more effects are included in the QED potential, then the Coulomb interactions, represented by the standard pair function, become insignificant. Then, this equation is identical to the Bethe–Salpeter equation. Solving the original BS equation iteratively, however, is extremely tedious and often very slowly converging, due to the dominating Coulomb interaction. As mentioned in the previous chapter, the QED-correlation equation is expected to be a faster road to reach the same goal. One- and two-photon exchange in the QED potential will very likely yield extremely good results, while such effects in the BS equation will often be quite insufficient (c.f. the discussion about the QED methods in Part II).

9.2 Quasi-Potential and Effective-Potential Approximations: Single-Reference Case

205

9.2 Quasi-Potential and Effective-Potential Approximations: Single-Reference Case In the equal-time approximation, where we equalize the times of the two particles in the Bethe–Salpeter equation (9.6), we can make a Fourier transformation of it with a single energy parameter as in the treatment of the single-particle Green’s function in Sect. 5.2.3. The Q part, falling outside the model space, then leads to: Q .E/ D Q G0 .E/ .i/˙ .E/ .E/

(9.15)

leaving out the space coordinates and integrations. Replacing the zeroth-order Green’s function with the resolvent (5.43) G0 .E/ D

i ; E H0

(9.16)

we obtain Q.E H0 / .E/ D Q˙ .E/ .E/:

(9.17)

If we identify the proper self-energy with the generalized potential (8.11) V.E/ D ˙ .E/;

(9.18)

the equation above leads together with the relation (6.125) P .H H0 /˝ .E/ D P V.E/ .E/

(9.19)

to The effective-potential form of the Bethe–Salpeter equation

.E H0 /j i D V.E/j i

(9.20)

frequently used in various applications. This equation was also derived above, using the Green’s operator only (6.126). The equation (9.20) can also be expressed j i Dj0 i C

Q V.E/j i ; E H0

(9.21)

where 0 is the model state 0 D P . This is equivalent to the Lippmann– Schwinger equation [12], frequently used in scattering theory. Formally, the equation (9.20) can also be expressed in the form of the time-independent Schrödinger equation H D E; (9.22)

206


where H is the energy-dependent Hamiltonian H.E/ D H0 C V.E/:

(9.23)

Equation (9.20) operates entirely in the restricted Hilbert space with constant number of photons. This can be related to the equivalent equation (6.32), derived by means of the Gell-Mann–Low theorem, which operates in the photonic Fock space. We can then regard the equation above as the projection of the Fock-space equation onto the restricted space.

9.3 Bethe–Salpeter–Bloch Equation: Multireference Case We can extend the treatment above to the general multireference case. From the expression (6.117), using the fact that the Green’s operator at time t D 0 is identical to the wave operator (6.52), we have in the single-reference case (one-dimensional) model space h iˇ ˛ j i D ˝j0 i D 1 C Q .E/ V.E/ C Q .E/ V.E/Q.E/ V .E/ C ˇ0 ; (9.24) wherej0 i is the model state,j0 i D P j i, and Q .E/ D

Q E H0

is the reduced resolvent (2.65). Operating on (9.24) from the left with Q.E H0 / now yields Q.E H0 /j i D QV.E/j i ;

(9.25)

which is identical to the equation (9.17) with the identification (9.18). For a general multidimensional (quasi-degenerate) model space, we have similarly Q.E ˛ H0 /j ˛ i D QV.E ˛ /j ˛ i (9.26) and This leads to or in operator form

P .E ˛ H0 /j ˛ i D P V.E ˛ /j ˛ i :

(9.27)

.E ˛ H0 /j ˛ i D V.E ˛ /j ˛ i

(9.28)

.Heff H0 /˝P D V.Heff /˝P

(9.29)

9.3 Bethe–Salpeter–Bloch Equation: Multireference Case

207

using the notations introduced in Sect. 6.9. But Heff ˝P D ˝Heff P D ˝H0 P C ˝Veff P;

which yields the commutator relation

/˝P ˝P Veff P; ˝; H0 P D V.Heff

(9.30)

/˝P . Here, the energy parameter of where according to (6.139) Veff P D P V.Heff V.Heff / is given by the model-space state to the far right, while the energy parameter of ˝ of the folded term depends on the intermediate model-space state (see footnote in Sect. 6.6). This equation is valid in the general multireference (quasidegenerate) situation and represents an extension of the effective-potential form (9.20) of the Bethe–Salpeter equation. Due to its close resemblance with the standard Bloch equation of MBPT (2.56), we refer to it as the Bethe–Salpeter–Bloch equation. This is equivalent to the generalized Bethe–Salpeter equation, derived in Chap. 6 (6.140). In analogy with the MBPT treatment in Sect. 2.5, we can separate the BS-Bloch equation into

˝1 ; H0 P D V.Heff /˝P ˝P Veff .H0 /P ; linked;1 /˝P ˝P Veff .H0 /P ; ˝2 ; H0 P D V.Heff linked;2

(9.31)

etc. It should be noted that the potential operator V.Heff / is an operator or matrix where each element is an operator/matrix. In the first iteration, we set Heff D H0 and .1/ in the next iteration Heff D H0 C Veff etc. Continued iterations correspond to the sum term in the expression (6.96), representing the model-space contributions. The two-particle BS-Bloch equation above is an extension of the ordinary pair equation, discussed in Sect. 2.5 (Fig. 2.6). The Bethe–Salpeter–Bloch equation leads to a perturbation expansion of Rayleigh–Schrödinger or linked-diagram type, analogous to that of standard MBPT expansions. It differs from the standard Bloch equation by the fact that the Coulomb interaction is replaced by all irreducible multiphoton interactions. Solving the BS-Bloch equation (9.30) is NOT equivalent to solving the singlestate equation for a number of states. The Bloch equation (9.30) leads to a Rayleigh/Schrödinger/linked-diagram expansion with folded terms that is size extensive. The single-state equation (9.20), on the other hand, leads to a Brillouin– Wigner expansion (see footnote in Sect. 2.4), that is not size extensive. Due to the very complicate form of the potential of the Bethe–Salpeter–Bloch equation, it is very difficult to handle this equation in its full extent. In the previous chapters, we have considered a simpler way of achieving essentially the same goal.

208


9.4 Problems with the Bethe–Salpeter Equation There are several fundamental problems with the Bethe–Salpeter equation and with relativistic quantum mechanics in general, as briefly mentioned in the Introduction. Dyson says in his 1953 paper [6] that this is a subject “full of obscurities and unsolved problems.” The question concerns the relation between the three-dimensional and the four-dimensional wave functions. In standard quantum mechanics, the three-dimensional wave function describes the system at a particular time, while the four-dimensional two-particle wave function describes the probability amplitude for finding particle one at a certain position at a certain time and particle two at another position at another time, etc. The latter view is that of the Bethe–Salpeter equation, and Dyson establishes a connection between the two views. The main problem is here the individual times associated with the particles involved, the physical meaning of which is not completely understood. This problem was further analyzed by Wick [24] and Cutkoski [4] and others. The relative time of the particles leads to a number of anomalous or spurious states – states that do not have nonrelativistic counterparts. This problem was analyzed in detail in 1965 by Nakanishi [13], and the situation was summarized in 1997 in a comprehensive paper by Namyslowski [14]. The Bethe–Salpeter equation was originally set up for the bound-state problem involving nucleons, such as the ground state of the deuteron. The equation has lately been extensively used for scattering problems in quantum chromodynamics, quark–quark/antiquark scattering. The equation has also been used for a long time in high-accuracy works on simple atomic systems, such as positronium, muonium, hydrogen, and helium-like ions. The problems with the BS equation, associated with the relative time, are most pronounced at strong coupling and assumed to be negligible in atomic physics, due to the very weak coupling. One important question is, of course, whether this is true also in the very high accuracy that is achieved in recent time. To attack the BS equation directly is very complicated, and for that reason various approximations and alternative schemes have been developed. The most obvious approximation is to eliminate the relative time of the particles, the equal-time approximation or external-potential approach. The first application of this technique seems to be have been made in the thesis of Sucher in the late 1950s [20, 21] for the evaluation of the leading QED corrections to the energy levels of the helium atom. This work has been extended by Douglas and Kroll [5] and by Drake, Zhang, and coworkers [25, 26], as will be further discussed in the Chap. 11. Another early application of an effective-potential approach was that of Grotch and Yennie [11] to obtain high-order effects of the nuclear recoil on the energy levels of atomic hydrogen. They derived an “effective potential” from scattering theory and applied that in a Schrödinger-like equation. A similar approach was applied to strongly interacting nucleons in the same year by Gross [10], assuming that one of the particles was “on the mass shell.” Related techniques have been applied to bound-state QED problems among others by Caswell and Lepage [2] and by Bodwin et al. [1]. A more

References

209

formal derivation of a “quasipotential” method for scattering as well as bound-state problems was made by Todorov [22], starting from the Lippmann–Schwinger scattering theory [12]. Several attempts have been made to correct for the equal-time approximation. Sazdjian [16, 17] has converted the BSE into two equations, one for the relative time and one eigenvalue equation of Schrödinger type. Connell [3] has developed a series of approximations, which ultimately are claimed to lead to the exact BSE. The approaches were primarily intended for strong interactions, but Connell tested the method on QED problems.

References 1. Boldwin, G.T., Yennie, D.R., Gregorio, M.A.: Recoil effects in the hyperfine structure of QED bound states. Rev. Mod. Phys. 57, 723–82 (1985) 2. Caswell, W.E., Lepage, G.P.: Reduction of the Bethe-Salpeter equation to an equivalent Schrödinger equation, with applications. Phys. Rev. A 18, 810–19 (1978) 3. Connell, J.H.: QED test of a Bethe-Salpeter solution method. Phys. Rev. D 43, 1393–1402 (1991) 4. Cutkosky, R.E.: Solutions of the Bethe-Salpeter equation. Phys. Rev. 96, 1135–41 (1954) 5. Douglas, M.H., Kroll, N.M.: Quantum Electrodynamical Corrections to the Fine Structure of Helium. Ann. Phys. (N.Y.) 82, 89–155 (1974) 6. Dyson, F.J.: The Wave Function of a Relativistic System. Phys. Rev. 91, 1543–50 (1953) 7. Feynman, R.P.: Space-Time Approach to Quantum Electrodynamics. Phys. Rev. 76, 769–88 (1949) 8. Feynman, R.P.: The Theory of Positrons. Phys. Rev. 76, 749–59 (1949) 9. Gell-Mann, M., Low, F.: Bound States in Quantum Field Theory. Phys. Rev. 84, 350–54 (1951) 10. Gross, F.: Three-dimensionsl Covariant Integral Equations for Low-Energy Systems. Phys. Rev. 186, 1448–62 (1969) 11. Grotch, H., Yennie, D.R.: Effective Potential Model for Calculating Nuclear Corrections to the Eenergy Levels of Hydrogen. Rev. Mod. Phys. 41, 350–74 (1969) 12. Lippmann, B.A., Schwinger, J.: Variational Principles for Scattering Processes. I. Phys. Rev. 79, 469–80 (1950) 13. Nakanishi, N.: Normalization condition and normal and abnormal solutions of Bethe-Salpeter equation. Phys. Rev. 138, B1182 (1965) 14. Namyslowski, J.M.: The Relativistic Bound State Wave Function. in Light-Front Quantization and Non-Perturbative QCD , J.P. Vary and F. Wolz, eds. (International Institute of Theoretical and Applied Physics, Ames) (1997) 15. Salpeter, E.E., Bethe, H.A.: A Relativistic Equation for Bound-State Problems. Phys. Rev. 84, 1232–42 (1951) 16. Sazdjian, H.: Relativistiv wave equations for the dynamics of two interacting particles. Phys. Rev. D 33, 3401–24 (1987) 17. Sazdjian, H.: The connection of two-particle relativistic quantum mechanics with the BetheSalpeter equation. J. Math. Phys. 28, 2618–38 (1987) 18. Schwinger, J.: On quantum electrodynamics and the magnetic moment of the electron. Phys. Rev. 73, 416 (1948) 19. Schwinger, J.: Quantum electrodynamics I. A covariant formulation. Phys. Rev. 74, 1439 (1948) 20. Sucher, J.: Energy Levels of the Two-Electron Atom to Order ˛ 3 Ry; Ionization Energy of Helium. Phys. Rev. 109, 1010–11 (1957)

210


21. Sucher, J.: Ph.D. thesis, Columbia University (1958). Univ. Microfilm Internat., Ann Arbor, Michigan 22. Todorov, I.T.: Quasipotential Equation Corresponding to the relativistic Eikonal Approximation. Phys. Rev. D 3, 2351–56 (1971) 23. Tomanaga, S.: On Infinite Field Reactions in Quantum Field Theory. Phys. Rev. 74, 224–25 (1948) 24. Wick, G.C.: Properties of Bethe-Salpeter Wave Functions. Phys. Rev. 96, 1124–34 (1954) 25. Zhang, T.: Corrections to O.˛ 7 .ln ˛/mc 2 / fine-structure splittings and O.˛ 6 .ln ˛/mc 2 / energy levels in helium. Phys. Rev. A 54, 1252–1312 (1996) 26. Zhang, T., Drake, G.W.F.: A rigorous treatment of O.˛ 6 mc 2 / QED corrections to the fine structure splitting of helium. J. Phys. B 27, L311–16 (1994)

Chapter 10

Implementation of the MBPT–QED Procedure with Numerical Results

In this chapter, we shall see how the combined covariant-evolution-QED approach developed in the previous chapters can be implemented numerically. In principle, this is equivalent to solving the complete Bethe–Salpeter equation perturbatively, but in practice, of course, approximations have to be made. We shall consistently work in the Coulomb gauge.1 We shall restrict ourselves here to the exchange of a single transverse photon together with a number of Coulomb interactions. We shall first apply the procedure in the no-pair case, and later a different procedure in the presence of virtual pairs will be applied. We work in the photonic Fock space, and initially we shall derive some relations for that space.

10.1 The Fock-Space Bloch Equation We have seen earlier that with the perturbation (6.35) H.x/ D H.t; x/ D O .x/ ec˛ A .x/ O .x/;

(10.1)

the wave function partly lies in an extended photonic Fock space, where the number of photons is no longer constant. According to the Gell-Mann–Low theorem, we have a Schrödinger-like equation (6.32) in that space .H0 C VF /j ˛ i D E ˛j ˛ i ;

(10.2)

where VF is the perturbation (6.36) with the Coulomb and the transverse parts, VF D VC C vT . We shall demonstrate below that, for single-photon exchange, this leads to a perturbation that is time-independent in the Schrödinger picture, which is a requirement for the GML theorem. Furthermore, in working in the extended space

1

This chapter is mainly based on [3, 5] and, in particular, on the thesis of Daniel Hedendahl [1].


211

212

10 Implementation of the MBPT–QED Procedure with Numerical Results

with uncontracted perturbations, it is necessary to include in the model Hamiltonian (H0 ) also the energy operator of the photon field (6.38) H0 D H0 C c ai .k/ ai .k/;

(10.3)

where D jkj. The wave operator is, as before, given by the Green’s operator at t D 0 (6.52), which may now contain uncontracted photon terms j ˛ i D ˝j0˛ i :

(10.4)

˛ j0˛ D P j ˛ i is the corresponding model state, which lies entirely within the restricted space with no uncontracted photons. From the GML equation (10.2), it can be shown in the same way as for the restricted space that the standard Bloch equation (2.56) is still valid also in the extended space

˝; H0 P D VF ˝ ˝Veff P:

(10.5)

The effective interaction is here given by Veff D P VF ˝P (6.61). The equation is formally the same as in the standard MBPT (2.56), but the operators involved now have somewhat different interpretation. The expressions for single transverse-photon exchange are given by (8.11). In the Coulomb gauge, these expressions involve the functions fTC , given by (4.60) fTC ./ D

e2 sin.r12 / sin.r12 / ˛ .˛ r / .˛ r / ˛ : 1 2 1 1 2 2 4 2 0 r12 2 r12

(10.6)

By means of the expansion theorem 1

X sin r12 D .2l C 1/jl .r1 /jl .r2 / C l .1/ C l .2/; r12

(10.7)

lD0

where jl .r/ are radial Bessel functions and C l vector spherical harmonics [4], we can express the function fTC as a sum of products of single-electron operators [5, Appendix B] fTC ./ D

1 h X lD0

i VGl .r1 / VGl .r2 / C Vsrl .r1 / Vsrl .r2 / ;

(10.8)

10.2 Single-Photon Potential in Coulomb Gauge: No Virtual Pairs

213

where VGl .r/ D Vsrl .r/ D

e p

2 0

r

e p

2 0

p

.2l C 1/ jl .r/ ˛C l ;

(10.9a)

hp .l C 1/.2l C 3/ jlC1 .r/ f˛C lC1 gl 2l C 1 i p (10.9b) C l.2l 1/ jl1 .r/ f˛C l1 gl :

In (10.6), the first term represents the Gaunt interaction and the second term the scalar retardation, which together form the Breit interaction (see Appendix F.2.2). Each term in the expansion – which are all time independent in the Schrödinger picture – will together with the Coulomb interaction (VC D e 2 =4 r12 ) form the (time-independent) perturbation VF D VC C VGl .r/ C Vsrl .r/:

(10.10)

10.2 Single-Photon Potential in Coulomb Gauge: No Virtual Pairs We consider first the case where no virtual pairs are present. Inserting the perturbation (10.10) into the Bloch equation (10.5) yields ˝; H0 P D VC C V l ˝P ˝ Veff ;

(10.11)

where we use V l as a short-hand notation for the Gaunt and scalar-retardation parts in (10.10). We consider first a number of instantaneous Coulomb interaction, forming a standard pair function (2.97), including also the zeroth order, ˝I PE D 1 C Q .E/I Pair PE :

(10.12)

This includes also the folds and is represented by the first diagram in Fig. 10.1. Then we can perturb this by one of the V l terms, representing part of the transverse Breit interaction, leading to the equation l ˝ ; H0 PE D V l ˝I PE ˝ l PE 0 I Pair PE

(10.13)

or E l E l D V l j˝Iab i j˝cd E h0 .1/ h0 .2/ c j˝ab hcd jI Pairjabi ;

(10.14)

214

10 Implementation of the MBPT–QED Procedure with Numerical Results r r r u

t a 6

E

E

6b a 6

E

6d

s

-

u t 6 c 6 E0

t

6b a 6

r

u

t

6b a 6

E

s

-

t 6 u c 6 E0

6b a 6

E

6u 6d 6b

Fig. 10.1 Expansion of the Bloch equation (10.11) in the Fock space, no virtual pairs, leading to single-photon exchange, including folded diagrams

(E; E 0 are here the energies of the unperturbed states jabi and jcd i, respectively). This equation has the solution E D ˇ ˇ l D ruˇ hruj˝ab

ˇ E Vl ˇ ˇ˝Iab E "r "u c ˇ D ˇ E˝ ˇ ˇ ˛ Vl ˇ ˇ ruˇ cd ˇI Pairˇab ; ˇ˝ Icd 0 .E "r "u c/.E "r "u c/

(10.15)

where ˝Ipq represents a pair function (10.12) starting from the statejpqi. Here, the first term is represented by the second diagrams in the Fig. 10.1 and the last term by the third “folded” diagram. The double bar indicates here the double denominator, which yields the first-order derivative (difference ratio) of the potential. By adding a second perturbation V l , we can complete the single-photon exchange between the electrons, which corresponds to solving the pair equation

E h0 .1/ h0 .2/ ˝sp PE D V l ˝ l PE ˝sp PE 0 I Pair PE :

(10.16)

This yields hrsj˝spjabi D

hrsjV l jrui hrujV l j˝Iab i .E "r "s /.E "r "u c/

hrsjV l jrui hrujV l j˝Icd i .E "r "s /.E "r "u c/E 0 "r "u c/

ˇ ˛ hrsj˝spjcd i hcd jI Pairjabi ˝ ˇ cd ˇI Pairˇab : E "r "s (10.17) This is illustrated by the last two diagrams of Fig. 10.1 (except for the final folded contribution).

10.2 Single-Photon Potential in Coulomb Gauge: No Virtual Pairs

215

Summing over and l, including the Gaunt as well as the scalar-retardation parts and considering photon emission from both electrons, we see that the result is in agreement with the Bloch equation (8.68) to first order ı.Q Vsp / PE 0 I Pair PE ıE ıVsp D Q .E/Vsp ˝I PE C Q .E/ PE 0 I Pair PE ıE Q .E/Q .E 0 /Vsp PE 0 I Pair PE

˝sp PE D Q .E/Vsp ˝I PE C

(10.18)

with (the first part of) the potential (6.16), ˇZ 1 ˇ c d fTC ./ hrsjVsp .E0 /jtui D rs ˇˇ 0 ˇ ˇ 1 1 ˇ tu (10.19) C E0 "r "u c E0 "t "s c ˇ (the second part of the potential is generated by emitting a photon from the second electron.) We see that the energy dependence of the potential is generated by an energy denominator

and the energy derivative (difference ratio) by a folded contribution (double denominator), when operating in the extended space (10.15). After the first interaction V1l , it is possible to add one or more Coulomb interactions, before closing the photon, corresponding to the first diagram in Fig. 10.2. This can be achieved by another iteration of the pair equation (10.14) E h0 .1/ h0 .2/ c ˝ l PE D VC ˝ l PE ˝ l PE 0 I Pair PE : (10.20) Instead of closing the uncontracted perturbation on the other electron, it can be closed on the same electron as the emission occurred from. This leads to a selfenergy interaction, represented by the first diagram in Fig. 10.3. This is a radiative effect (see Sect. 2.6), which is infinite and has to be renormalized, as discussed in Chap. 12. 0

r0 6 r 6

6u

E

6b

6 -

6u

t 6 a 6

6

6

6 6

6

Fig. 10.2 Crossing Coulomb interactions before closing the retarded interaction (left). Continuing the process leads to the single transverse-photon exchange combined with high-order electron correlation, including crossing Coulomb interactions (right)

216

10 Implementation of the MBPT–QED Procedure with Numerical Results r k’ k

r

u v

v t a 6

r

6b

a

k

s

u

t

E

k’

u

t

E

b

a

E

b

Fig. 10.3 Second-order contribution to the wave operator in the extended Fock space

A second perturbation V l can also be applied without contracting the first one, leading to diagrams indicated by the second and third diagrams in Fig. 10.3. Closing these photons lead to irreducible two-photon interactions, discussed in Sect. 8.2.

10.3 Single-Photon Exchange: Virtual Pairs 10.3.1 Illustration The iterative procedure of the previous section works well in the no-pair situation, when the repeated single-photon exchange leads to reducible diagrams of ladder type, which means that they can in time-ordered form be separated into legitimate diagrams by horizontal cuts. In the presence of virtual pairs, we have to use a different procedure. As we have seen above (Sect. 8.3), we have to combine the general potential with Coulomb interactions (8.67) to be able to treat the potential in an iterative process. This will at the same time eliminate the so-called Brown–Ravenhall effect of vanishing energy denominators. This potential (8.67) can be used directly in the Bloch equation (8.68). In principle, we can use the corresponding full potential with QED effects (8.67), but for simplicity we shall consider only the pure single-photon part. We use pair functions (10.12) as input, and perturbing this with the potential (8.67) leads in next order to

ı.Q V QED / ˝sp P D Q V QED ˝I C ˝I PI Pair P: ıE

(10.21)

For the evaluation we use, as before, the expansion (10.8) 1 ˝ ˇ C ˇ ˛ X rs ˇfT ./ˇtu D hsjVGl .r2 /jui hrjVGl .r1 /jti lD0

C hsjVsrl .r2 /jui hrjVsrl .r1 /jti :

(10.22)

10.3 Single-Photon Exchange: Virtual Pairs

217

r 6

Usp

-

f t 6VP 6 a 6

6s Vsp 6u VC 6 ˝Iab 6b

As an illustration we consider the second term in Fig. 8.11 (shown above), when there is a single hole (t) [c.f. (8.13)]. Then (10.21) becomes

ı.Q Usp / Pair ˝sp P D Q Usp ˝I C P; ˝I PI ıE

(10.23)

where Usp D Vsp Q VC represents the transverse photon (Vsp ) with a Coulomb interaction. Here, only the first and the third terms of the potential Vsp (8.11) are relevant, yielding for the first term jrC sC i E " r "s

"

hsC jV l juC i hrC jV l jt i hsC jV l juC i hrC jV l jt i C "t "r c E "r "u c

ht uC jVCj˝Iab i ; E "t "u

#

(10.24)

where again V l represents the Gaunt and the scalar-retardation potentials for the two electrons. In order to evaluate the expression above, we first perform an additional iteration of the pair equation (10.12) ˙˛ E h0 .1/ h0 .2/ j˝ab D VCj˝Iab i Q VCj˝Icd i hcd jI Pairjabi

(10.25)

yielding a new pair function with a single hole output. The solution can be expressed in analogy with (10.15) ˝ ˇ Pairˇ ˛ ˛ ht uC jVCj˝Iab i ht uC jVCj˝Icd i ˙ cd ˇI ˇab : D ht uCj˝ab E "t " u .E "t "u /.E 0 "t "u / (10.26) The first part of the solution, illustrated in Fig. 10.4 (a), represents the last factor in the expression (10.24). The second folded part will be used later in constructing the complete folded contribution. In evaluating the first term within the square brackets of (10.24), we first multiply the pair function (10.26) without folded contribution by

hrC jV1l jt i "t "r c

218


a

c

b u 6 6

a6

t

6 b 6

r Vl t 6 a6

u 6

r

6

t 6

b 6

a6

6

s 6 u 6

d

s 6 u

r

6

t 6

b 6

a6

6

6 b 6

Fig. 10.4 (a–c): Generating a pair function with a hole output (10.25), combined with a singleparticle perturbation V1 , and closed with a perturbation V2 . The last diagram contains a crossing Coulomb interaction, as discussed at the end of the section

yielding l iD hrC uC j˝ab

˙ i hrC jV1l jt iht uC j˝ab hrC jV1l jt i ht uC jVCj˝Iab i D "t "r c "t "r c E "t "u

and represented by the diagram (b) in Fig. 10.4. Then we close the photon by multiplying with hsC jV2l juC i and including the final denominator, yielding hrC sC j˝spjabi D

hsC jV2l juC i hrC jV1l jt i ht uC jVCj˝Iab i ; E "r "s "t "r c E "t "u

(10.27)

which agrees with the corresponding part of (10.24) (Fig. 10.4c). The second term in the brackets of (10.24) is evaluated in a similar way with a different denominator. The folded contribution is in lowest order from (10.21) ı.Q Usp / ı.Q Vsp Q VC / j˝Icd i hcd jI Pairjabi D j˝Icd i hcd jI Pairjabi : ıE ıE (10.28) Here, ıVsp ı.Q Vsp Q VC / ı.Q VC / ıQ D Vsp Q VC C Q Q VC C Q Vsp : (10.29) ıE ıE ıE ıE The difference ratio ı.Q VC /=ıE is obtained from the folded part of (10.26) (VC is energy independent). Similarly, the difference ratio of the relevant part of Vsp is obtained by including an extra factor

E0

1 : "r "u c

Finally, the difference ratio of Q is ıQ 1 D 0 : ıE .E "r "s /.E "r "s /

10.3 Single-Photon Exchange: Virtual Pairs

219

It should be noted that the last term in (10.29) is combined with the pair function (10.26) with folded contribution, while the other terms are combined with that function without that contribution. This is important to avoid the singularities of Brown–Ravenhall type, mentioned above.

10.3.2 Full Treatment We shall now generalize the treatment above and consider all 16 combinations of the single-photon exchange (8.11) (see Fig. 8.3) essentially in one single step. To begin with, we leave out the folded contribution. Then, the expression to evaluate is hrjV l jti hsjV l jui htujVCj˝Iab i E " r "s t˙ r t˙ s˙ u˙ r ˙ u ˙ s ˙ : ˙ ˙ ˙ "t "r ˙ c E "t "s c E "r "u c "u "s ˙ c (10.30) If we in this expression make the substitution t r $ us, we get an identical result but with a and b interchanged. Therefore, we can replace the expression above by the much simpler expression hrjV l jti hsjV l jui htujVCj˝Iab C ˝Iba i E " r "s u˙ r˙ t˙ r : ˙ ˙ "t "r ˙ c E "r "u c

(10.31)

The last expression can be evaluated in the following way. We first evaluate the matrix element htujVCj˝Iab i by iterating the pair equation (10.25) once, allowing negative-energy states as output. We can separate the solutions into four block, depending on the signs of the outgoing orbital energies, as illustrated in the matrix in Fig. 10.5. Next, we evaluate the matrix elements hrjV1jti for each value of and l, and separate them in a similar way, shown in Fig. 10.6. We now multiply the matrices in Figs. 10.5 and 10.6 (in that order), leading to the matrix in Fig. 10.7. Here, we include the two denominator terms in the brackets of (10.31) and sum over all t, particle as well as hole states. Finally, we multiply the result by hsjV2jui and sum over and l, corresponding to closing the photon (c.f. Fig. 10.4c), and apply the final denominator in (10.31). If the input orbitals a; b are different, the procedure is repeated with a $ b. The folded contribution in (10.21) is evaluated in a similar way [c.f. (10.28)].

220

10 Implementation of the MBPT–QED Procedure with Numerical Results 1

0

Fig. 10.5 Representation of the pair function in (10.25) iterated one extra time and separated into four blocks, depending on the signs of the outgoing orbital energies

B B B B B B B B B B B B B B B B B B B B B B B B B @

6t

u 6

6t

6

6

6

6

a 6

b 6

a 6

b 6

u 6 u

6

t

6

6

t

6

a 6

b 6

a 6

b 6

0

Fig. 10.6 The matrix elements hrjV1jt i, separated in analogy with Fig. 10.5

Fig. 10.7 Result of multiplying the matrices in Figs. 10.5 and 10.6. The t line represents particle as well as hole states

u

1

B B B r B 6V B 1 B B B B t 6 B B B B B B B B B B B B B B Br B t @

t

6r

t ? r ?

0 B B B B B B B B B B B B B B B B B B B B B B B B B @

C C C C C C C C C C C C C C C C C C C C C C C C C A


1 r 6 t˙ 6

r

6u

6 a 6

6 b 6

t˙ 6

6u

6 a 6

6 b 6

r 6 t˙ 6

6 a6

r

6u b 6

t˙ 6

6 a6

6u b 6


10.4 Numerical Results

221

6 6

6 -

6

6

6

6

6

Fig. 10.8 The Feynman diagrams representing a single transverse photon exchange combined with high-order electron correlation (heavy horizontal line). The internal vertical lines represent electron propagators with particle and holes. The numerical evaluation of this diagram is given below

10.3.3 Higher Orders In most of the cases treated above, it is possible also to insert Coulomb interaction before the photon interaction is completed, as discussed in the no-pair case. This is the case when the orbitals u; r or t; s are of the same kind (particle or hole), as indicated in Fig. 10.4d. This corresponds to including another part of the potential in (8.66). After the completion of the single-photon exchange, the iteration process can be continued with further Coulomb interactions, leading to the complete single-photon exchange with electron correlation, including all combinations of particles and holes, as illustrated in Fig. 10.8.

10.4 Numerical Results 10.4.1 Two-Photon Exchange In Chap. 7 (Fig. 7.3), we showed the results of two-photon exchange for the groundstate of helium-like ions, calculated using the S-matrix formulation. In Table 10.1, we compare these results with those obtained by Hedendahl et al. [1, 3], in testing the new covariant method described in the present chapter. The agreement, which is found to be very good, is also displayed in Fig. 10.9, where the solid lines represent the old S-matrix results and the squares the new covariant method. As before, the scale is logarithmic and the norm is the nonrelativistic ionization energy.

10.4.2 Beyond Two Photons Calculations have also been performed of the effect of electron correlation beyond two-photon exchange for the ground state of helium-like ions by Hedendahl et al. [1–3]. Some results are shown in Table 10.2 and also displayed in Fig. 10.10.

222

10 Implementation of the MBPT–QED Procedure with Numerical Results Table 10.1 Comparisons between two-photon effects for He-like ions ground states, evaluated with the S-matrix and the covariant-evolution-operator methods (in Hartree) (see Fig. 7.3) Coul.–Breit Coul.–Breit. Coul.–Breit Z Method NVPA retard. VP uncrossed. 6 S-matrix 1; 054:2 31:4 10:1 6 CEO 1; 054:9 31:5 10:0 10 S-matrix 2; 070:4 122:3 45:9 10 CEO 2; 071:0 122:4 45:9 14 S-matrix 5; 515 292:8 121:5 14 CEO 5; 517 292:8 121:2 18 S-matrix 8; 947 553:1 247:3 18 CEO 8; 950 553:3 248:2 30 CEO 23; 632 1; 909:9 1; 010

Fig. 10.9 Comparison of some two-photon exchange contributions (Coulomb–Breit NVPA, Coulomb–Breit retardation, and Coulomb–Breit virtual pairs, no correlation) for the ground-state of some helium-like ions obtained by S-matrix calculations (see Fig. 7.3) (heavy lines) and by means of the covariant-evolution procedure (squares), described in this chapter (see Table 10.1, c.f. Fig. 7.3 in Chap. 7) (from [1, 3])

The top line of the figure, representing the Coulomb–Breit interaction with correlation without virtual pairs, contains the instantaneous as well as the retarded Breit interaction. The former part lies within the no-virtual-pair approximation (NVPA) and is therefore NOT a QED effect with the definition we have previously made. In order to obtain the pure QED effect, the instantaneous part is subtracted, yielding the retarded part, represented by the second line of the figure. The next line represents the same effect with Coulomb crossings, and the bottom line represents the effect of electron correlation on the Coulomb–Breit interaction with virtual pairs and no

10.4 Numerical Results Table 10.2 Contributions due to electron correlation beyond two-photon exchange for the ground state of some helium-like ions. This can be compared with the corresponding two-photon exchange in Table 10.1 (in H)

223

Z 6 10 14 18 30 42

Beyond two-photon Coul.–Breit Unretarded Retarded Virt.pairs 137 17 2.7 223 40 7.3 301 68 13 372 100 21 553 210 46 688 322 71

Fig. 10.10 The effect of electron correlation beyond two-photon exchange – Coulomb–Breit NVPA, Coulomb–Breit retardation with and without Coulomb crossings, and Coulomb–Breit virtual pairs, all WITH electron correlation, for the ground-state of helium-like ions (c.f. Fig. 10.9) (from [1, 3]). For comparison, the effect of pure retarded two-photon exchange without additional correlation is also indicated

crossing Coulomb interactions. The corresponding Feynman diagrams are shown at the bottom of the figure. This represents the first numerical bound-state calculation beyond two-photon exchange. In the figure, we have for comparison also indicated the effect due to doubly retarded two-photon interactions (thin black line), estimated from the S-matrix results. This comparison demonstrates the important result that – starting from single-photon exchange – for light and medium-heavy elements the effect of electron correlation is much more important than a second retarded photon interaction.

224


10.4.3 Outlook The results presented here are incomplete and represent only the nonradiative part of the QED effect in combination with electron correlation. The corresponding radiative effects with a single transverse photon are also possible to evaluate. Such calculations are underway by the Gothenburg group. The effect due to double transverse photons is presently beyond reach, but the effect can be estimated by replacing the second transverse photon by an instantaneous Breit interaction. The calculations performed so far with the procedure described here concern the ground states of helium-like ions [3]. By extending the calculations to excited states, it will be possible to make detailed comparison with experimental data. For instance, very accurate data exist for some helium-like ions, as shown in Tables 7.7 and 7.8. In some of these cases, the experimental results are at least two orders of magnitude more accurate than the best theoretical estimates made so far. Furthermore, it seems that standard procedures applied until now cannot be significantly improved in this respect, so – to be able to make significant progress – there might be a need for a new, improved procedure, like the MBPT–QED procedure presented here. Then, it might be possible for the first time to observe the combined effect of QED and electron correlation.

References 1. Hedendahl, D.: Towards a Relativistic Covariant Many-Body Perturbation Theory. Ph.D. thesis, University of Gothenburg, Gothenburg, Sweden (2010) 2. Hedendahl, D., Lindgren, I., Salomonson, S.: Towards numerical implementation of the relativistically covariant many-body perturbation theory. Can. J. Phys. 87, 817–24 (2008) 3. Hedendahl, D., Salomonson, S., Lindgren, I.: . Phys. Rev. p. (to be published) (2011) 4. Lindgren, I., Morrison, J.: Atomic Many-Body Theory. Second edition, Springer-Verlag, Berlin (1986, reprinted 2009) 5. Lindgren, I., Salomonson, S., Hedendahl, D.: Many-body procedure for energy-dependent perturbation: Merging many-body perturbation theory with QED. Phys. Rev. A 73, 062,502 (2006)

Chapter 11

Analytical Treatment of the Bethe–Salpeter Equation

11.1 Helium Fine Structure The leading contributions to the helium fine structure beyond the first-order relativistic contribution (NVPA, see, Sect. 2.6) were first derived in 1957 by Araki [1] and Sucher [6, 7], starting from the Bethe–Salpeter (BS) equation [5] and including the nonrelativistic as well as the relativistic momentum regions. Following the approach of Sucher, Douglas and Kroll [2] have derived all terms of order ˛ 4 H(artree)1 , where no contributions in the relativistic region were found. The same approach was later used by Zhang [8, 12] to derive corrections of order ˛5 log ˛ H and of order ˛5 H in the nonrelativistic region and recoil corrections to order ˛ 4 m=M H (see also [10]). Later some additional effects of order ˛5 H due to relativistic momenta were found by Zhang and Drake [11]. The radiative parts are treated more rigorously by Zhang in a separate paper [9]. Using a different approach, Pachucki and Sapirstein [4] have derived all contributions of order ˛5 H and reported some disagreement with the early results of Zhang [8].2 We shall here follow the approach of Sucher in his thesis [7]. This is based directly on the BS equation, which makes it possible to identify the contributions in terms of Feynman diagrams and therefore to compare them with the results obtained in the previous chapters. This approach of Sucher is closely followed by Douglas and Kroll [2] and by Zhang [8], and we shall in our presentation make frequent references to the corresponding equations of Sucher (S), Douglas and Kroll (DK), and Zhang (Z).

1 H(artree) is the energy unit of the Hartree atomic unit system (see Appendix K.1). In the relativistic unit system, the energy unit is mc 2 D ˛ 2 H . 2 This chapter is largely based on the paper [3].


225

226

11 Analytical Treatment of the Bethe–Salpeter Equation

11.2 The Approach of Sucher The treatment of Sucher starts from the Bethe–Salpeter equation (9.5), which in our notations (9.6) reads, leaving out the inhomogeneous term (S 1.1, DK 2.5), .x; x 0 / D

ZZZZ

d4 x1 d4 x2 d4 x10 d4 x20

G00 .x; x 0 I x2 ; x20 / .i/˙ .x2 ; x20 I x1 ; x10 / .x1 ; x10 /:

(11.1)

G00 is the zeroth-order two-particle Green’s function, dressed with all kinds of single-particle self-energies. ˙ is identical to the irreducible potential V (Fig. 6.6). The undressed zeroth-order Green’s function is, using the relation (5.38), G0 .x; x 0 I x2 ; x20 / D G0 .x; x2 / G0 .x 0 ; x20 / D iSF .x; x2 / iSF .x 0 ; x20 /

(11.2)

and the corresponding dressed function is then G00 .x; x 0 I x2 ; x20 / D G.x; x2 / G.x 0 ; x20 / D iSF0 .x; x2 / iSF0 .x 0 ; x20 /;

(11.3)

where G is the full single-particle Green’s function, generated in the field of the nucleus (Furry representation) (see Fig. 5.1), and SF0 the correspondingly dressed electron propagator. The Green’s functions satisfy the relation (5.36) (S 1.5) @ i h1 G.x; x0 / D iı 4 .x x0 /; @t

(11.4)

which leads to (S 1.6, DK 2.19) i

@ h1 @t

ZZZZ @ i 0 h2 .x; x 0 / D i d4 x1 d4 x2 d4 x10 d4 x20 ı 4 .x x2 / @t ı 4 .x 0 x20 / ˙ .x2 ; x20 I x1 ; x10 / .x1 ; x10 / ZZ Di d4 x1 d4 x10 ˙ .x; x 0 I x1 ; x10 / .x1 ; x10 /; (11.5)

where h1;2 are the Dirac single-electron Hamiltonians for electron 1 and 2. We assume that the wave function is of the form .x; x 0 / D .T; ; x; x 0 / D eiE T .; x; x 0 /;

(11.6)

where T D .t C t 0 /=2 is the average time and D t t 0 is the relative time. Then @ @ 0 .x; x 0 /; i .x; x / D E=2 C i @t @ @ @ 0 .x; x 0 / i 0 .x; x / D E=2 i @t @

11.2 The Approach of Sucher

227

leading to (S 1.9, DK 2.23) @ @ E=2 C i E=2 i h1 h2 .; x; x 0 / @ @ Z ZZ D i d1 d3 x 1 d3 x 01 ˙ .; x; x 0 I 1 ; x 1 ; x 01 / .1 ; x 1 ; x 01 /(11.7) leaving out the average time. Sucher then transfers to the momentum representation, but we shall here still work in the coordinate representation with a Fourier transform only of the time variables. We define the Fourier transform with respect to time Z F ./ D d ei F ./ (11.8) and the inverse transformation Z F ./ D

d i F ./: e 2

(11.9)

Fourier transforming (11.7) with respect to yields .E=2 C h1 / .E=2 h2 / .; x 1 ; x 01 / ZZ Z d3 x 1 d3 x 01 ˙ .; x; x 0 I 1 ; x 1 ; x 01 / .1 ; x 1 ; x 01 /: (11.10) D i d1 Performing the Fourier transform of the right-hand side with respect to 1 yields Z

ZZ

d10 d1 i0 1 i1 1 ˙ .; x; x 0 I 10 ; x 1 ; x 01 / .1 ; x 1 ; x 01 / e 1 e d1 2 2 ZZ d10 d1 D 2ı.1 C10 / ˙ .; x; x 0 I 10 ; x 1 ; x 01 / .1 ; x 1 ; x 01 / (11.11) 2 2

or (S 1.16) .E=2 C h1 / .E=2 h2 / .; x 1 ; x 01 / Z ZZ d1 Di d3 x 1 d3 x 01 ˙ .; x; x 0 I 1 ; x 1 ; x 01 / .1 ; x 1 ; x 01 /: (11.12) 2 Following Sucher, we express the relation (11.10) in operator form FO j i D gO j i : The operator FO has the (diagonal) coordinate representation ˝ ˛ ; x; x 0 jFj; x; x 0 D .E=2 C h1 / .E=2 h2 /

(11.13)

(11.14)

228


and the operator gO has the (nondiagonal) representation ˝ ˛ ˛ i ˝ O 1 ; x 1 ; x 01 D ; x; x 0 jgj ; x; x 0 j˙O j1 ; x 1 ; x 01 : 2

(11.15)

We expand the interaction into gO D gO c C gO ;

(11.16)

where gO c represents the Columbic part of gO gO c D

i O Ic 2

(11.17)

and IOc is the Coulomb interaction with the (diagonal) coordinate representation ˛ ˝ ; x; x 0 jIOc j; x; x 0 D

e2 ; 4jx x1 j

(11.18)

gO represents the remaining part of gO gO D gO T C gT c C gO T c 2 C gO T T C C gO rad ;

(11.19)

where gO T represents a single transverse photon, gO T c and gO T c 2 a transverse photon with one and two crossing Coulomb interactions, respectively, gO T T with two irreducible transverse photons, and finally gO rad all radiative corrections. This corresponds to the diagrams are shown in Fig. 11.1. With the decomposition (11.16), the relation (11.13) becomes (S 1.30, DK 3.6) 1 j i D FO gO gO c j i

(11.20)

with the coordinate representation 1 ˝ ˛ j2 ; x 2 ; x 02 h; x; x 0j i D ; x; x 0 j FO gO ˛ ˝ 2 ; x 2 ; x 02 jgO c j1 ; x 1 ; x 01 h1 ; x 1 ; x 01j i

+

+

+

+

+

Fig. 11.1 Diagrammatic representation of the approximation in (11.19), used by Sucher

(11.21)

+

11.2 The Approach of Sucher

229

or noting that the representation of gO c is diagonal 1 ˛ ˝ h; x; x 0j i D ; x; x 0 j FO gO j1 ; x 1 ; x 01 gO c h1 ; x 1 ; x 01j i : Sucher defines the equal-time wave function (S 1.32, DK 3.8) Z ˚.x; x 0 / D d .; x; x 0 /

(11.22)

(11.23)

or in operator form j˚i Dji hj i ;

(11.24)

which gives with (11.22) 1 ˛ ˝ h; x; x 0j i D ; x; x 0 j FO gO jx 1 ; x 01 gO c hx 1 ; x 01j˚i :

(11.25)

Summing over with the replacement (11.17), this can be expressed as (S 1.34) Z j˚i D i

1 d O F gO IOc j˚i : 2

(11.26)

Using the identity (S 1.35, DK 3.11) .A B/1 A1 C A1 B.A B/1 ; the BSE (11.26) becomes (DK 3.12) Z i d h O 1 F C FO 1 gO .FO gO /1 IOc j˚i : j˚i D i 2

(11.27)

(11.28)

The inverse of the operator OF is FO 1 D

1

1 ; O E=2 C h1 E=2 hO 2

(11.29)

which is a product of electron propagators in operator form (4.14) FO 1 D SOF .E=2 C / SOF .E=2 /:

(11.30)

In the coordinate representation (4.12), SF .!I x; x 0 / D

hxjj i hj jx 0 i hxjj i hj jx 0 i hxjj i hj jx 0 i D C C : ! "j C i sgn."j / ! "j C i ! "j i (11.31)

230


Integration over then yields (S 1.44, DK 3.24) Z hx; x 0jrsi hrsjx0 ; x00 i d 1 .CC / ; F D i 2 E " r "s

(11.32)

which is also the negative of the Fourier transform of the zeroth-order Green’s function G0 .EI x; x 0 I x 0 ; x 00 /, or in operator form Z i d O 1 .CC / : (11.33) F D G0 .E/ D 2 E hO 1 hO 2 Equation (11.28) then becomes (S 1.47, DK 3.26)3 Z d 1 h1 C h2 C CC Ic C D i F g .F g /1 Ic ˚ D E ˚; 2 (11.34) where D D E h 1 h2 :

(11.35)

This is the starting point for the further analysis. The operator on the left-hand side can be written in the form Hc C H , where Hc D h1 C h2 C CC Ic CC

(11.36)

is the Hamiltonian of the no-(virtual-) pair Dirac-Coulomb equation (Z 16) Hc c D Ec c and

Z H D CC Ic .1 CC / Ic C D i D H1 C H2

(11.37)

d 1 F g .F g /1 Ic 2 (11.38)

is the remaining “QED part” (S 2.3, DK 3.29, Z 17). The first part H1 represents virtual pairs due to the Coulomb interaction and the second part effects of transverse photons (Breit interaction). In order to include electron self-energy and vacuum polarizations, the electron propagators (5.37) are replaced by propagators with self-energy insertions ˙./, properly renormalized (DK 2.10), S 0 ./ D

jri hrj : "r C ˇ˙./ C ir

(11.39)

Also renormalized photon self-energies have to be inserted into the photon lines. 3

In the following, we leave out the hat symbol on the operators.

11.3 Perturbation Expansion of the BS Equation

231

11.3 Perturbation Expansion of the BS Equation The effect of the QED Hamiltonian (11.38) can be expanded perturbatively, using the Brillouin–Wigner perturbation theory, E D E Ec D hc jV C V V C V V V C jc i V jc i ; D hc j 1 V

(11.40)

where is the reduced resolvent (2.65) D Q .E/ D

Q 1 jc i hc j 1 jc i hc j D D E Hc E Hc Dc

(11.41)

with Dc D E Hc :

(11.42)

The unperturbed wave function is in our case one solution of the no-pair Dirac– Coulomb equation (11.37), c , and we can assume that the perturbation is expanded in other eigenfunctions of Hc . Q is the projection operator that excludes the state c (assuming no degeneracy). This leads to the expansion (S 2 19–21, DK 3.43, Z 28) E .1/ D hc jHjc i ;

(11.43a)

E .2/ D hc jH Hjc i ;

(11.43b)

E .3/ D hc jH H Hjc i ;

(11.43c)

etc. Since CCjc i D jc i and jc i D 0, it follows that hc jH1jc i 0, and the first-order correction becomes (DK 3.44) Z d 1 .1/ E D hc jH2jc i D hc jD i (11.44) F J F 1 Icjc i 2 and (DK 3.45) J D g .1 F 1 g /1 :

(11.45)

The second-order corrections are (DK 3.46)4 Ea.2/ D hc jH1 H1jc i D hc jIc Ic jc i ;

(11.46a)

4 Note that the two Ic in (11.46a) are missing from [2, Eq. 3.46]. Equation (11.46b) agrees with [8, Eq.3.30] but not with [2], where the factor Ic LCC should be removed.

232


Z d 1 Eb.2/ D hc H1 H2 i c D c Ic D i F J F 1 Ic c ; 2 (11.46b)

Z d 1 Ec.2/ D hc H2 H1 i c D c D i F J F 1 Ic Ic c ; 2 (11.46c) Ed.2/ D hc H2 H2 i c

Z Z d 1 d 1 D c D i F J F 1 Ic D i F J F 1 Ic c : (11.46d) 2 2 These formulas can be simplified, noting that D D

Q .E h1 h2 /; E Hc

(11.47)

which, using the relation (11.42), becomes (DK 3.41) CC Ic CC D D 1 C D : E Hc

(11.48)

According to DK Ea.2/ ; Ec.2/ and E .3/ do not contribute to the fs in order ˛ (Hartree). This holds also in the next order according to Zhang, but E .3/ will contribute to the singlet energy in that order. In the relativistic momentum region, the second-order part Ea.2/ contributes to the energy already in order ˛ 3 H and to the fine structure in order ˛5 H [8, p. 1256]. Using the relation (11.42), we have Ec Hc D Dc CC Dc CC , and the no-pair equation (11.37) can be written (DK 3.51) 4

.Dc CC Ic / c D 0:

(11.49)

.2/

Then the second-order correction Eb (11.46b) can be expressed as Z

.2/

Eb D hc j.Ic Dc / i

d 1 F J F 1 Ic jc i : 2

(11.50)

This can be combined with the first-order correction E .1/ (11.44), yielding Z hc j.Ic C E/ i

d 1 F J F 1 Ic jc i 2

(11.51)

11.4 Diagrammatic Representation

233

with E D E Ec D D Dc :

(11.52)

Here, the E term differs in sign from (DK 3.54) and (Z 37). The reason for the discrepancy between our result here and those of DK and Z seems to be that the latter make the replacement (DK 3.48) 1 F 1 D S1 S2 .S1 C S2 / S11 C S21 D

S1 C S 2 D D 1 .S1 C S2 / ; E h1 h2 (11.53)

which follows from (11.30), and then approximate D with Dc in the second-order expression.

11.4 Diagrammatic Representation To continue, we make the expansion (DK 3.45, Z 32) J D g .1 F 1 g /1 D g C g F 1 g C ;

(11.54)

where the first term represents irreducible terms and the remaining ones are reducible. Furthermore, we make the separation (DK 3.53, Z 12) g D gT C g;

(11.55)

where gT represents the interaction of a single transverse photon and g the irreducible multiphoton exchange of (11.19). The first-order expression (11.44) becomes Z

d 1 .1/ F gT C gT F 1 gT C g C F 1 Ic jc i (11.56) E D hc jD i 2 and the leading terms are illustrated in Fig. 11.2. The first term can be expanded in no-pair and virtual-pair terms (a–c) E

.1/

Z D hc jD i

d 1 F gT F 1 .CC CC CC C /Ic jc i : (11.57) 2

The second term in (11.56) represents in lowest order two reducible transverse photons (d) and the third term irreducible (inclusive radiative) multiphoton part, (e–h).

234


a

c

b

d

g

e

f

h

Fig. 11.2 Diagrammatic representation of the first-order expression (11.56)

a

c

b

d

Fig. 11.3 Diagrammatic representation of the second-order expressions (11.58a)–(11.58d)

Similarly, the second-order expressions above become Ea.2/ D hc jIc Ic jc i ; .2/ Eb

(11.58a)

Z

1 d 1 1 F gT C gT F gT C g C F Ic c ; D c Ic i 2 (11.58b)

Z

d 1 Ec.2/ D c i F gT C gT F 1 gT C g C F 1 Ic Ic c ; 2 (11.58c) Ed.2/

Z

d 1 F ŒgT C F 1 Ic 2 Z d 1 1 D i F ŒgT C F Ic c : 2

D c D i

(11.58d)

This is illustrated in Fig. 11.3. The first second-order contribution (11.58a) represents two Coulomb interactions with double pair (Fig. 11.3a) and the next

References

235

contribution (11.58b) in lowest order a transverse photon and a Coulomb interaction with double pair (b). The third contribution represents in lowest order one transverse photon and two Coulomb interactions with a double pair (c). The last term represents two reducible transverse photons with at least one Coulomb interaction (d).

11.5 Comparison with the Numerical Approach In the previous chapter, we have described an approach that is presently being developed by the Gothenburg group of treating the Bethe–Salpeter equation numerically. This is based on the covariant-evolution approach and the Green’s-operator technique, described previously, and to a large extent upon the numerical techniques developed by the group and applied to numerous atomic systems (see Sect. 2.7). This new technique has the advantage over the analytical approach that all relativistic effects are automatically included in the procedure. This simplifies the handling appreciably, and it corresponds to the treatment of the entire Sect. 4 of Douglas and Kroll [2] or to Sect. VII in the paper of Zhang [8]. The numerical technique of solving the Bethe–Salpeter equation, described in the previous chapter, is presently only partly developed, but the effect of one transverse photon with arbitrary number of crossing Coulomb interactions can presently be handled as well as virtual pairs. This corresponds to most of the terms gT CgT c C gT c 2 C of the expansion in (11.19) and to the numerous formulas of Sect. 5 of Douglas–Kroll and of Sect. IV of Zhang. Also part of the multiphoton effect can be treated numerically by iterating reducible interactions with a single transverse photon, corresponding to the operator gT F gT in the formulas above with crossing Coulomb interactions. These effects are treated in Sect. 6 of Douglas and Kroll. The irreducible interaction with several transverse photons cannot be teated at present with the numerical technique, but this can be approximated with one retarded and one or several unretarded photons (instantaneous Breit). Also radiative effects can be handled with the same approximation.

References 1. Araki, H.: Quantum-Electrodynamical Corrections to Energy-levels of Helium. Prog. Theor. Phys. (Japan) 17, 619–42 (1957) 2. Douglas, M.H., Kroll, N.M.: Quantum Electrodynamical Corrections to the Fine Structure of Helium. Ann. Phys. (N.Y.) 82, 89–155 (1974) 3. Lindgren, I.: The helium fine-structure controversy. arXiv;quant-ph p. 0810.0823v (2008) 4. Pachucki, K., Sapirstein, J.: Contributions to helium fine structure of order m˛7 . J. Phys. B 33, 5297–5305 (2000) 5. Salpeter, E.E., Bethe, H.A.: A Relativistic Equation for Bound-State Problems. Phys. Rev. 84, 1232–42 (1951)

236


6. Sucher, J.: Energy Levels of the Two-Electron Atom to Order ˛ 3 Ry; Ionization Energy of Helium. Phys. Rev. 109, 1010–11 (1957) 7. Sucher, J.: Ph.D. thesis, Columbia University (1958). Univ. Microfilm Internat., Ann Arbor, Michigan 8. Zhang, T.: Corrections to O.˛ 7 .ln ˛/mc 2 / fine-structure splittings and O.˛ 6 .ln ˛/mc 2 / energy levels in helium. Phys. Rev. A 54, 1252–1312 (1996) 9. Zhang, T.: QED corrections to O.˛ 7 mc 2 / fine-structure splitting in helium. Phys. Rev. A 53, 3896–3914 (1996) 10. Zhang, T.: Three-body corrections to O.˛ 6 / fine-structure in helium. Phys. Rev. A 56, 270–77 (1997) 11. Zhang, T., Drake, G.W.F.: Corrections to O.˛ 7 mc 2 / fine-structure splitting in helium. Phys. Rev. A 54, 4882–4922 (1996) 12. Zhang, T., Yan, Z.C., Drake, G.W.F.: QED Corrections of O.mc 2 ˛ 7 ln ˛/ to the Fine Structure Splitting of Helium and He-like Ions. Phys. Rev. Lett. 77, 1715–18 (1996)

Chapter 12

Regularization and Renormalization

(See, for instance, Mandl and Shaw [17, Chap. 9] and Peskin and Schroeder [23, Chap. 7].) In previous chapters, we have evaluated some radiative effects in the S-matrix (Chap. 4) and covariant-evolution operator formulations (Chap. 8). In this chapter, we discuss the important processes of renormalization and regularization in some detail. Many integrals appearing in QED are divergent, and these divergences can be removed by replacing the bare electron mass and charge by the corresponding physical quantities. Since infinities are involved, this process of renormalization is a delicate matter. In order to do this properly, the integrals first have to be regularized, which implies that the integrals are modified so that they become finite. This has to be done so that the process is gauge-independent. After renormalization, the regularization modification is removed. Several regularization schemes have been developed, and we shall consider some of them in this chapter. If the procedure is performed properly, the way of regularization should have no effect on the final result.

12.1 The Free-Electron QED 12.1.1 The Free-Electron Propagator The wave functions for free electrons are given by (D.29) in Appendix D pC .x/ D .2/3=2 uC .p/ eip x eiEp t ; p .x/ D .2/3=2 u .p/ eip x eiEp t

(12.1)

states (r D 1; 2) where p is the momentum vector and pC represents positive-energy p and p negative-energy states (r D 3; 4). Ep D cp0 D c 2 p2 C m2 c 4 . The coordinate representation of the free-electron propagator (4.10) then becomes hx1 jSOFfreejx2 i D

Z

d! X p;r .x 1 / p;r .x 2 / i!.t1 t2 / e ; 2 p;r ! "free p .1 i/


(12.2)

237

238

12 Regularization and Renormalization

free where "free p is the energy eigenvalue of the free-electron function (Ep D j"p j). The Fourier transform with respect to time then becomes X p;r .x 1 / p;r .x 2 / ) free .1 i/ ! " p p;r Z eip.x1 x 2 / d3 p X u .p/ u .p/ D r r .2/3 r ! "free p .1 i/ Z 3 d p 1 uC .p/ uC .p/ D 3 .2/ ! Ep .1 i/ 1 C u .p/ u .p/ eip.x1 x2 / : ! C Ep .1 i/

hx1 jSOFfreejx2 i D

The square bracket above is the Fourier transform of the propagator, and using the relations (D.35, D.36), this becomes1 SFfree .!; p/ D

1 1 1 C 2 ! Ep .1 i/ ! C Ep .1 i/ 2 1 1 c ˛ p C ˇmc C 2p0 ! Ep .1 i/ ! C Ep .1 i/

! C c ˛ p C ˇmc 2 ! C c˛ p C ˇmc 2 D ! 2 Ep2 C i ! 2 .c 2 p2 m2 c 4 / .1 i/ 1 D ! .c ˛ p C ˇmc 2 / .1 i/

D

(12.3)

with Ep2 D c 2 p02 D c 2 p2 C m2 c 4 and ˛ˇ D ˇ˛. This can also be expressed SFfree .!; p/ D

1 !

hfree D .p/ .1

i/

;

(12.4)

where hfree D .p/ is the momentum representation of the free-electron Dirac Hamiltonian operator (D.21), hO free O D .p/. Formally, we can write (12.3) inp covariant four-component form with ! D cp0 with cp0 disconnected from Ep D c 2 p2 C m2 c 4 – known as off the mass-shell. Then we have2 1 SFfree .p/ D cSFfree .!; p/ D ˇ 6 p mc C i

1

(12.5)

In the following, we shall for simplicity denote the electron physical mass by m instead of me . The factor of ˇ appears here because we define the electron propagator (4.9) by means of O instead of the more conventionally used NO D O ˇ. 2

12.1 The Free-Electron QED

239

with 6 p D p D ˇ˛ p D ˇ.p0 ˛p/ D .p0 C˛p/ ˇ (see Appendix D). Note that the two transforms differ by a factor of c (c.f. Sect. 4.3, see also Appendix K).

12.1.2 The Free-Electron Self-Energy The S-matrix for the first-order free-electron self-energy (Fig. 12.1) is obtained from (4.84) and (4.44) with the momentum functions (12.1) after time integrations S .2/ .!I p0 ; r 0 ; p; r/ D e 2 c 2

Z

dz 2

ZZ

0

d3 x d3 x 0 ur0 .p0 / eip x

0

˛ iSFfree .! zI x 0 ; x/ ˛ ur .p/ eipx iDF .z; x 0 x/: (12.6) The relation between the momentum and coordinate representations is Z

d3 q free 0 SF .!; q/ eiq.x x/ ; 3 .2/ Z d3 k 0 free DF .z; k/ eik.x x/ : DF .zI x 0 ; x/ D 3 .2/ SFfree .!I x 0 ; x/

D

(12.7) (12.8)

Integration over the space coordinates then yields Z d3 q d3 k 3 S .!I p ; r ; p; r/ D e c ı .p q k/ ı 3 .p0 q k/ .2/3 .2/3 Z dz free free ur0 .p0 / .z; k/ ur .p/ ˛ SF .! z; k/ ˛ DF 2 (12.9) and integration over q 0

.2/

Z

0

2 2

S .2/ .!I p0 ; r 0 ; p; r/ D ı 3 .p0 p/ ur0 .p0 / .i/˙ free .!; p/ ur .p/;

(12.10)

where Z ˙

free

.!; p/ D ie c

2 2

dz 2

Z

d3 k free ˛ SF .! z; k/ ˛ DF .z; k/: .2/3 ! p0 r 0 6 qs

Fig. 12.1 Diagram representing the first-order free-electron self-energy

6

pr

! 6

zk

240


In covariant notation we have, using z D ck0 , Z ˙ free .p/ D ie 2 c 2

d4 k free ˛ SF .p k/ ˛ DF .k/; .2/4

(12.11)

which is the free-electron self-energy function. With the expression (12.5) for the free-electron propagator, this becomes expressed in terms of gamma matrices Z ˇ˙

free

6 p 6 k C mc d4 k DF .k/ .2/4 .p k/2 m2 c 2 C i

.p/ D ie c

2 2

(12.12)

or Z ˇ˙ free .p/ D ie 2 c 2

1 d4 k DF .k/: 4 .2/ 6 p 6 k mc C i

(12.13)

With the commutation rules in Appendix (D.58), this becomes Z ˇ˙

free

6 p 6 k 2mc d4 k DF .k/ .2/4 .p k/2 m2 c 2 C i

.p/ D 2ie c

2 2

(12.14)

and with the photon propagator (4.28) we have in the Feynman gauge ˇ˙ free .p/ D

2ie 2 c 0

Z

6 p 6 k 2mc d4 k 1 : 4 2 2 2 2 .2/ .p k/ m c C i k C i

(12.15)

fAs mentioned above, the factor of ˇ is due to our definition of the electron propagator [c.f. (12.5)]g.

12.1.3 The Free-Electron Vertex Correction We consider first the single interaction with an external energy potential (Appendix D.41) e˛ A (Fig. 12.2 left). The S-matrix is given by S .1/ .! 0 ; !I p0 r 0 ; pr; q/ D iec

Z

0

d3 x ur0 .p0 / eip x ˛ A .x/ ur .p/ eipx

(12.16)

S .1/ .! 0 ; !I p0 r 0 ; pr; q/ D iec ı 3 .p p0 / ur 0 .p0 / ˛ A .p p0 / ur .p/;

(12.17)

or

where A .q/ is the Fourier transform of A .x/.

12.1 The Free-Electron QED

241

p0 r 0 6 A .q/

A .q00 /

pr

6

! p0 r 0 6 q0 6

6q pr

zk

! 6

Fig. 12.2 Diagram representing the first-order free-electron vertex correction

The vertex-modified free-electron self-energy diagram in Fig. 12.2 (right) becomes similarly ZZZ Z dz 0 0 d3 x 1 d3 x 2 d3 x 3 ur 0 .p0 / eip x S .3/ .! 0 ; !I p0 r 0 ; pr/ D .ie/3 c 2 2 ˛ iSFfree .! 0 z; x 0 ; x 00 /˛ A .x 00 / ˛ iSFfree .! z; x 00 ; x/ ur .p/ eipx iDF .z; x 0 x/: (12.18) In analogy with (12.9), this becomes Z Z Z 3 00 Z d3 q d3 q0 d q d3 k 0 u 0 .p / S .3/ .! 0 ; !I p0 r 0 ; pr/ D e 3 c 2 3 3 3 .2/ .2/ .2/ .2/3 r ı 3 .p q k/ ı 3 .p0 q0 k/ ı 3 .q q0 C q00 / Z dz free 0 ˛ SF .! z; q0 / 2 ˛ A .q00 / ˛ SFfree .! z; q/ ur .p/ DF .z; k/ (12.19) 0

and after integrations over q, q , and q

00

S .3/ .! 0 ; !I p0 r 0 ; pr/ D ie ı 3 .p p0 / ur 0 .p0 / .! 0 ; !I p0 ; p/ A .p p0 / ur .p/;

(12.20) where 0

0

Z

.! ; !I p ; p/ D ie c

2 2

dz 2

Z

d3 k free 0 ˛ SF .! z; p0 k/ .2/3

˛ SFfree .! z; p k/ ˛ DF .z; k/

(12.21)

is the vertex correction function. In covariant notations, this becomes in analogy with (12.10) Z d4 k free 0 0 2 .p ; p/ D ie c ˛ SF .p k/ ˛ SFfree .p k/ ˛ DF .k/ (12.22) .2/4

242


In the Feynman gauge, this becomes ie 2 0

Z

1 d4 k 0 4 .2/ 6 p 6 k mc C i 1 1 2 : 6 p 6 k mc C i k C i

.p 0 ; p/ D

(12.23)

Comparing with the self-energy function (12.15), we find the Ward identity (4.100) [17, Eq. 9.60] @ ˙.p/ D .p; p/: (12.24) @cp Obviously, this relation holds independently of the gauge.

12.2 Renormalization Process We shall here derive expressions for the mass and charge renormalization in terms of counterterms that can be applied in evaluating the QED effects on bound states. The process of regularization will be treated in the next section.

12.2.1 Mass Renormalization We consider now a bare electron with the mass m0 . The corresponding free-electron propagator (12.5) is then SF bare .!; p/ D ˇ

1 6 p c m0 c 2 C i

(12.25)

with ! D cp0 . We now “dress” the bare-electron propagator with all kinds of self-energy insertions in the same way as for the bound-electron propagator in Fig. 5.7. This corresponds to the S-matrix in operator form3 iSF .!; p/ C iSF .!; p/.i/˙.!; p/ iSF .!; p/ C D

3

iSF .!; p/ ; 1 ˙.!; p/ SF .!; p/ (12.26)

Note that ˙.!; p/ has the dimension of energy and that the product ˙.!; p/ SF .!; p/ is dimensionless (see Appendix K).

12.2 Renormalization Process

D

243

C

C

C D

C

Fig. 12.3 Dyson equation for the dressed bare-mass electron propagator

D

˙ free .p/=

C 6

C 6

C

Fig. 12.4 Expansion of the proper self-energy operator for a bare electron

which leads to 1 SF bare;dressed .!; p/ D ˇ 2 6 p c m0 c ˇ˙bare .!; p/ C i

(12.27)

illustrated in Fig. 12.3. Here, the box represents the irreducible or proper self-energy insertions, ˙bare .!; p/, illustrated in Fig. 12.4. We shall in the following refer to this as the free-electron self-energy, ˙ free .!; p/, .!; p/ D ˙ free .!; p/: ˙bare

(12.28)

To lowest order, the free-electron self-energy is in analogy with (4.85) Z ˙

free

.!; p/ D i

d! bare S .!; p/ I bare .!I p/; 2 F

(12.29)

where I bare is the interaction (4.44) in the momentum representation with the electronic charge replaced by the bare charge, e0 . The bare-electron propagator itself is also associated with a bare-electron charge (e0 ) at each vertex. The dressing of the electron propagator leads to a modification of the electron mass as well as of the electron charge. One part of the free-electron self-energy is indistinguishable from the mass term in the electron propagator, and another part is indistinguishable from the electronic charge, and these parts give rise to the mass renormalization and the charge renormalization, respectively. The modification of the electron charge is here compensated by a corresponding modification of the vertex (to be discussed below), so that there is no net effect on the electron charge in connection with the electron self-energy. On the other hand, there is a real modification of the electron charge in connection with the modification of the photon propagator, as we shall discuss later.

244


Instead of working with the bare-electron mass and charge with self-energy insertions, we can use the physical mass and charge and introduce corresponding counterterms (see, for instance, [13, p. 332]). The free-electron propagator with the physical electron mass, m, is 1 SF free .!; p/ D ˇ 6 p c mc 2 C i

(12.30)

and it has its poles “on the mass shell,” 6 p D mc [see Appendix (D.19)]. The dressed propagator (12.27) should have the same pole positions, which leads with

to

m D m0 C ım

(12.31)

ˇ ımc 2 D ˇ˙ free .!; p/ˇ6p Dmc :

(12.32)

This is the mass-counterterm. We can now in the dressed operator (12.27) replace m0 c 2 by ˇ mc 2 ˇ˙ free .!; p/ˇ6p Dmc ; which leads to

where

1 ; SF free;ren .!; p/ D ˇ 2 free .!; p/ C i 6 p c mc ˇ˙ren

(12.33)

ˇ free ˙ren .!; p/ D ˙ free .!; p/ ˙ free .!; p/ˇ6p Dmc :

(12.34)

This represents the mass-renormalization. Both the free-electron self-energy and the mass counterterms are divergent, while the renormalized self-energy is finite.

12.2.2 Charge Renormalization 12.2.2.1 Electron Self-Energy The pole values (residues) of the dressed bare electron propagator should also be the same as for the physical propagator, including the associated electronic charges. The physical propagator (12.30) with the electronic charge e 2 SFfree .!; p/ D ˇ

e2 6 p c mc 2 C i


245

has the pole value ˇe 2 =c. The dressed propagator (12.27) with the bare electron charge is ˇ

e02 e02 D ˇ free .!; p/ C i 6 p c m0 c 2 ˇ˙ free .!; p/ C i 6 p c mc 2 ˇ˙ren

and its pole value at the pole6 p D mc is lim

6p !mc

ˇ e02 .6 p mc/ e02 ˇ D ˇ free .!; p/ C i @ free .!; p/ˇ c 6 p c mc 2 ˇ˙ren c 1 ˇ @c6p ˙ren C i 6p Dmc

using l’Hospital’s rule. This gives us the relation e2 D

e02

ˇ @ free .!; p/ˇ 1 ˇ @c6p ˙ren 6p Dmc

(12.35)

or ˇ @ ˇ free .!; p/ ˇ : ˙ren e 2 D e02 1 C ˇ 6p Dmc @c 6 p

(12.36)

Here, the second term, which is divergent, represents the first-order charge renormalization. It is convenient to express the free-electron self-energy as ˙ free .!; p/ D A C B.6 p c mc 2 / C C.6 p c mc 2 /2 :

(12.37)

It then follows that the constant A is associated with the mass renormalization, ˇ A D ˙ free .!; p/ˇ6p Dmc D ˇımc 2

(12.38)

and B with the charge renormalization, BD

ˇ @ ˙ free .!; p/ˇ6p Dmc : @c 6 p

(12.39)

From (12.36), it follows that for the charge renormalization due to the dressing of the electron propagator becomes e D e0 .1 C B=2 C /:

(12.40)

The constant C represents the renormalized free-electron self-energy that is finite.

246


12.2.2.2 Vertex Correction The modification of the vertex function shown in Fig. 12.2 can be represented by ie0 .p; p 0 / D ie0 ˛ ie0 ˇ .p; p 0 /;

(12.41)

where e0 is the “bare” electron charge. The vertex correction is divergent and can be separated into a divergent part and a renormalized, finite part 0 .p; p 0 / D L˛ C ren .p; p /:

(12.42)

The divergent vertex part corresponds to a charge renormalization, in first order being e D e0 .1 ˇL/: (12.43) But this should be combined with the charge renormalization due to the dressing of the electron propagators (12.40), which yields e D e0 .1 ˇL C ˇB/;

(12.44)

since there are two propagators associated with each vertex. Due to the Ward identity (12.24), it then follows that the charge renormalization due to the electron selfenergy and the vertex correction exactly cancel. This holds also in higher orders. 12.2.2.3 Photon Self-Energy We first transform the first-order photon self-energy (4.108) to the momentum representation, using Z

d4 k .2/4 Z d4 k DF .x4 ; x2 / D .2/4 Z d4 q SF .x3 ; x4 / D .2/4 Z d4 q SF .x4 ; x3 / D .2/4 DF .x1 ; x3 / D

eik.x1 x3 / DF .k/; 0

eik .x4 x2 / DF .k 0 /; eiq.x3 x4 / SF .q/; 0

eiq .x4 x3 / SF .q 0 /:

(12.45)

The space integrations over x3 and x3 give rise to the delta functions ı 4 .k q C q 0 / and ı 4 .k 0 q C q 0 /, yielding with the bare electron charge e02 , Z

d4 k ie 2 ˛ DF .k/ i˘3;4 .k/ ie02 ˛2 DF .k/ .2/4 0 1 Z d4 q Tr i˛3 SF .q/ i ˛4 SF .q k/ : i˘3;4 .k/ D .2/4

(12.46)


247

1

2

+

1

2

3

2

+

Fig. 12.5 Diagram representing the first-order vacuum polarization of the single photon (firstorder photon self-energy)

The photon self-energy represents a modification of the single-photon exchange, illustrated in Fig. 12.5, ie02 DF .k/ ) ie02 DF .k/ C ie02 DF .k/ i˘ .k/ ie02 DF .k/ C : (12.47) With the form (4.28) of the photon propagator in the Feynman gauge, this becomes ie02 g ie02 g ie02 g ie02 g ) C i˘ .k/ : 2 2 2 c 0 k C i c 0 k C i c 0 k C i c 0 k 2 C i (12.48) From the Lorentz covariance, it follows that the polarization tensor must have the form ˘ .k/ D g A.k 2 / C k k B.k 2 /

(12.49)

and it can be shown that in this case only the second term can contribute [4, p. 155], [17, p. 184]. This reduces the expression above to ie02 g e02 A.k 2 / ie02 g 1 ) c 0 k 2 C i c 0 k 2 C i c 0 k 2 C i

ie02 c 0

k2 C

g : A.k 2 / C i

e02 c0

(12.50)

The expression above represents the modification of the photon propagator due to the photon self-energy. It is infinite and can be interpreted as a change of the electronic charge – or charge renormalization – in analogy with the mass renormalization treated above. The photon propagator has a pole at k 2 D 0, corresponding to the zero photon mass [c.f. the free-electron propagator in (12.5)], and the pole value is proportional to the electron charge squared, e0 2. If A.k 2 D 0/ D 0;

(12.51)

248


then also the modified propagator has a pole at the same place with a pole value proportional to e02 : (12.52) 2 2 ˇ e 1 C 0 dA.k2 / ˇ 2 c0

dk

k D0

This cannot be distinguished from the bare charge and represents the physical electron charge, e02 e02 dA.k 2 / ˇˇ 2 2 e0 1 ; (12.53) e D 2 2 ˇ e2 c 0 d k 2 k D0 1 C 0 dA.k2 / ˇ 2 c0

dk

k D0

which is the charge renormalization. The polarization tensor may have a finite part that vanishes at k 2 D 0, ˘ren , which is the renormalized photon self-energy. This is physically observable. 12.2.2.4 Higher-Order Renormalization The procedure described above for the first-order renormalization can be extended to higher orders. A second-order procedure has been described by Labzowsky and Mitrushenkov [14] and by Lindgren et al. [15], but we shall not be concerned with that further here.

12.3 Bound-State Renormalization: Cutoff Procedures Before applying the renormalization procedure, the divergent integrals have to modified so that they become finite, which is the regularization procedure. Details of this process depend strongly on the gauge used. Essentially, all QED calculations performed so far have been carried out in the so-called covariant gauges (see Appendix G), preferably the Feynman gauge. In the remaining sections of this chapter, we shall review some of the procedures used in that gauge, and also consider the question of regularization in the Coulomb gauge. Several regularization procedures have been developed, and the conceptually simplest ones are the cutoff procedures. The most well known of these procedure is that of Pauli–Willars and another is the so-called partial-wave regularization. A more general and more sophisticated procedure is the dimensional regularization, which has definite advantages and is frequently used today. We shall consider some of these processes in the following sections.

12.3.1 Mass Renormalization When we express the Dirac Hamiltonian (2.108) with the physical mass „D D c˛ pO C ˇmc 2 C vext ;

(12.54)

12.3 Bound-State Renormalization: Cutoff Procedures Fig. 12.6 Diagram representing the renormalization of the first-order self-energy of a bound electron

249 r 2 t 1 a

-

r 6

a 6

we have to include the mass counterterm (12.32) in the perturbation density (6.35) H.x/ D ec O .x/ ˛ A .x/ O .x/ ımc 2 O .x/ˇ O .x/:

(12.55)

The bound-electron self-energy operator is given by (8.47) ˇ

ˇZ ˇ dz bou ˇ iSF ."a zI x 2 ; x 1 / I.zI x2 ; x 1 /ˇˇ ta hrj˙ bou ."a /jai D rt ˇˇ 2

(12.56)

and subtracting the corresponding mass-counterterm yields the renormalized selfenergy operator ˇ ˛ ˝ ˇ bou ."a /jai D r ˇ˙ bou ."a / ˇımc 2 ˇ a hrj˙ren

(12.57)

illustrated in Fig. 12.6. Here, both terms contain singularities, which have to be eliminated, which is the regularization process. In the regularization process due to Pauli and Villars [21], [17, Eq. 9.21], the following replacement is made in the photon propagator 1 1 1 ) 2 ; k 2 C i k 2 C i k 2 2 C i

(12.58)

which cuts off the ultraviolet and possible infrared divergence.

12.3.2 Evaluation of the Mass Term (See Mandl and Shaw [17, Sect. 10.2].) On the mass shell, p 6 D mc, the free-electron self-energy (12.15) becomes [17, Eq. 10.16] ˇ 2e 2 ım c D ˙ free .p/ j6pDmc D i c 0 2

Z

6 k C mc 1 d4 k : (12.59) 4 2 2 .2/ k 2pk C i k C i

250


In order to evaluate this integral, we apply the Pauli–Villars regularization scheme, which we can express as 1 1 1 ) 2 D k 2 C i k 2 C i k 2 2 C i

Z

2

2

dt : .k 2 t C i/2

(12.60)

By means of the identity (J.4) in Appendix J with a D k 2 t and b D k 2 2pk we can express the mass term ım c 2 D

4ie 2 0

Z

Z

d4 k .2/4

Z

2

1

dt

dx

2

0

Œk 2

.6 k C mc/x : 2pk.1 x/ tx3

(12.61)

With the substitutions q D p.1 x/ and s D tx, the k integral becomes, using the integral (J.8) and (J.9) and 6 p D mc, Z

d4 k .6 k C mc/x mc x.2 x/ i D 4 2 3 2 2 .2/ Œk C 2qp C s 32 m c 2 .1 x/2 C tx

(12.62)

yielding ımc 2 D

e 2 mc 8 2 0

Z

1

dx .2 x/ ln

0

2 x C m2 c 2 .1 x/2 : 2 x C m2 c 2 .1 x/2

(12.63)

This is logarithmically divergent as ! 1 with the leading term being e 2 mc ımc D 8 2 0

Z

1

2

0

2 x : dx .2 x/ ln 2 C ln m .1 x/2

(12.64)

To evaluate the second part of the integral, we need the following formulas Z

Z dx ln x D x ln x x

which leads to

Z

1

dx x ln x D

x 2 ln x x 2 ; 2 4

(12.65)

x 3 D : (12.66) 2 .1 x/ 4 0 ı In all unit systems with „ D 1, the factor e 2 4 0 D c ˛, where ˛ is the finestructure constant (see Appendix K), and the mass term (12.59) becomes I D

dx .2 x/ ln

3˛ mc 2 ım./ c D 2 2

1 ln C : mc 4

(12.67)

12.3 Bound-State Renormalization: Cutoff Procedures

251

12.3.3 Bethe’s Nonrelativistic Treatment Bethe’s original nonrelativistic treatment of the Lamb shift [3] is of great historical interest, and it also gives some valuable insight into the physical process. Therefore, we shall briefly summarize it here. From the relation (4.90), we have the bound-state self-energy, using the Feynman gauge (4.55), hx 2 j˙

bou

e2c ."a /jx 1 i D hx 2 j˛jti 4 0 r12

Z

1

0

d sin r12 htj˛ jx 1 i; "a "t c sgn"t (12.68)

where r12 D jx1 x 2 j. For small k values and positive intermediate states, this reduces to Z 1 d e2c bou (12.69) ˙ ."a / D 2 ˛jti htj˛ : 4 0 " "t c a 0 The scalar part of ˛ ˛ cancels in the renormalization, leaving only the vector part to be considered, ˙ bou ."a / D

e2c ˛jti 4 2 0

Z

1

d htj˛: "a "t c

0

(12.70)

The corresponding operator for a free electron in the state pC is ˙ free .pC / D

˛ e2c ˛jqC 2 4 0

Z

1 0

˝ d q j˛ "qC c C

"pC

(12.71)

restricting the intermediate states to positive energies. In the momentum representation, this becomes ˝

ˇ ˛ ˇ ˛ e2 c ˝ 0 p0C ˇ˙ free .pC /ˇpC D pC j˛jqC 2 4 0

Z 0

1

"pC

˝ ˛ d qC j˛jpC : "qC c (12.72)

But since ˛ is diagonal with respect to the momentum, we must have q D p D p0 . Thus, ˝

ˇ ˛ ˇ p0C ˇ˙ free .pC /ˇpC D ıp30 ;p

˛ˇ2 e 2 ˇˇ˝ pC j˛jpC ˇ 2 4 0

Z

1

d:

(12.73)

0

Obviously, this quantity is infinite. Inserting a set of complete states, this becomes ˝

ˇ

ˇ

p0C ˇ˙ free .pC /ˇpC

˛

D

ıp30 ;p

˛ e2 ˝ p j˛jti htj˛jpC 4 2 0 C

Z

1

d: 0

(12.74)

252


The free-electron self-energy operator can then be expressed as ˙ free .pC / D ıp30 ;p

e2 ˛jti 4 2 0

Z

1

d htj˛;

(12.75)

0

which should be subtracted from the bound-electron self-energy operator (12.70). We can assume the intermediate states ftg to be identical to those in the bound case. This gives the renormalized self-energy operator Z 1 e2 "a "t bou ."a / D ˛jti d (12.76) ˙ren htj˛ : 2 4 0 "a "t c 0 The expectation value of this operator in a bound statejai yields the renormalized bound-electron self-energy in this approximation, i.e., the corresponding contribution to the physical Lamb shift, Z 1 e2 "a "t bou ."a / jai D d : (12.77) haj˛jti htj˛jai haj ˙ren 4 2 0 r12 " "t c a 0 This result is derived in a covariant Feynman gauge, where the quantized radiation has transverse as well as longitudinal components. In the Coulomb gauge, only the former are quantized. Since all three vector components above yield the same contribution, we will get the result in the Coulomb gauge by multiplying by 2/3. Furthermore, in the nonrelativistic limit, we have ˛ ! p=c, which leads to bou ."a / jai D haj ˙ren

e2 hajpjti htjpjai 2 6 c 2 0 r12

Z

1

d 0

"a "t ; (12.78) "a "t c

which is essentially the result of Bethe. Numerically, Bethe obtained the value 1,040 MHz for the shift in the first excited state of the hydrogen atom, which is very close to the value 1,000 MHz obtained experimentally by Lamb and Retherford. Later, the experimental shift has been determined to be about 1,057 MHz. Bethe’s results was, of course, partly fortuitous, considering the approximations made. However, it was the first success performance of a renormalization procedure and represented a breakthrough in the theory of QED. We can note that the nonrelativistic treatment leads to a linear divergence of the self-energy, while the relativistic treatment above gives only a logarithmic divergence.

12.3.4 Brown–Langer–Schaefer Regularization The bound-state electron propagator can be expanded into a zero-potential term, a one-potential term, and a many-potential term


a 2

a 6 2

t

= t6

1 a

1 a 6

+ x

253

a 6 2 6 3

6 1 a 6

x

+ x

a 6 2 6 4 3

16 a 6

Fig. 12.7 Expanding the bound-state self-energy in free-electron states according to (4.107)

SFbou .!; x 2 ; x 1 / D SFfree .!; x 2 ; x 1 / Z C d3 x 3 SFfree .!; x 2 ; x 3 / v.x 3 /SFfree .!; x 3 ; x 1 / Z C d3 x 3 d3 x 4 SFfree .!; x 2 ; x 4 / v.x 4 / SFbou .!; x 4 ; x 3 / v.x 3 /SFfree .!; x 3 ; x 1 /;

(12.79)

which leads to the expansion of the bound-electron self-energy, as illustrated in Fig. 12.7, ˇ E D ˇ Z dz ˇ ˇ SFfree ."a zI x 2 ; x 1 / I.zI x 2 ; x 1 / ˇta haj˙ bou ."a /jai D at ˇ 2 Z D ˇZ dz free ˇ S ."a zI x 2 ; x 3 / v.x 3 / C at ˇ d3 x 3 2 F ˇ ED ˇ Z ˇ ˇ CSFfree ."a zI x 3 ; x 1 / I.zI x 2 ; x 1 / ˇta at ˇ d3 x 3 d3 x 4 Z Z dz d3 x 3 d3 x 4 SFfree .!; x 2 ; x 4 / v.x4 / 2 ˇ E ˇ SFbou .!; x 4 ; x 3 / v.x3 /SFfree .!; x 3 ; x 1 / I.zI x 2 ; x 1 / ˇta ; (12.80) where I.zI x 2 ; x 1 / represents the single-photon interaction (4.45). We can then express this as haj˙ bou ."a /jai D haj˙ free ."a /jai hajecA free ."a /jai C haj˙mpjai : (12.81) Here, the first term on the right-hand side is the average of the free-electron self-energy in the bound state jai, the second term a vertex correction (4.98) with v.x/ D e˛ A , and the last term the “many-potential” term.

254


We can now use the expansion (12.37) of the free-electron self-energy in (12.81), where the first term (A) will be eliminated by the mass-counterterm in (12.57). We are then left with the average of the mass-renormalized free-electron self-energy (12.34), which is still charge divergent. If we separate the vertex operator in a divergent and a renormalized part according to (12.42), it follows from (12.44) that the charge-divergent parts cancel, and we are left with three finite contributions, the mass-renormalized free-electron self-energy (12.34), the many-potential term (12.79), and the finite part of the vertex correction (12.42) bou free ."a /jai D hrj˙ren ."a /jai hrjeA free;ren ."a /jai C hrj˙mpjai : hrj˙ren (12.82)

This is the method of Brown et al. [6], introduced already in 1959. It was first applied by Brown and Mayers [7] and later by Desidero and Johnson [10], Cheng et al. [8,9], and others. The problem in applying this expression lies in the many-potential term, but Blundell and Snyderman [5] have devised a method of evaluating this terms numerically with high accuracy (and the remaining terms analytically). We can also express the renormalized, bound self-energy (12.57) as bou ."a /jai D hrj˙ bou ."a /jai hrj˙ free ."a /jai hrj˙ren C hrj˙ free ."a /jai hrjˇımc 2jai ;

(12.83)

where the second term is the renormalized free-electron self-energy (12.34), evaluated between bound states. This is illustrated in Fig. 12.8. The mass term can be evaluated by expanding the bound states in momentum representation ˝ (12.84) hrjˇımc 2jai D hrjp0 ; r 0 i p0 ; r 0 j˙ free ."p /jp; ri hp; rjai as illustrated in Fig. 12.9. The relation (12.83) can then be written as ˇ ˛ ˝ ˇ ˝ ˇ ˝ bou ."a /jai D r ˇ˙ bou ."a / ˙ free ."a /ˇa C r ˇp0 ; r 0 i p0 ; r 0 j˙ free ."a / hrj˙ren ˇ ˛ ˙ free ."p /jp; ri hp; r ˇa ; (12.85) where we note that in the last term the energy parameter of the self-energy operator is equal to the energy of the free particle.

r 2

(

t 1 a

-

r 6 2 t 6

r 6 2

)

+

1 a 6

Fig. 12.8 Illustration of the method of Brown et al.

(

t 6 1 a 6

-

r 6

a 6

)


r 6

255

p0 ; r 0 6 2

=

hrjp0 ; r 0i

q; s 1 p; r

a 6

6

z; k

hp; rjai

6

Fig. 12.9 Expansion of the mass term in momentum space

In this way, the leading mass-divergence term is eliminated, while the parts are still charge-divergent, but this divergence is cancelled between the parts. The elimination of the mass-renormalization improves the numerical accuracy. Mohr has developed the method further and included also the one-potential part of the expansion in the two parts, thereby eliminating also the charge divergence. In this way, very accurate self-energies have been evaluated for hydrogenic systems [18–20].

12.3.5 Partial-Wave Regularization An alternative scheme for regularizing the electron self-energy is known as the partial-wave regularization (PWR), introduced independently by the Gothenburg and Oxford groups [16, 24]. 12.3.5.1 Feynman Gauge Using the expansion (10.7) 1

X sin r12 D .2l C 1/jl .r1 /jl .r2 / C l .1/ C l .2/; r12

(12.86)

lD0

the expression (4.90) for the bound-state self-energy in the Feynman gauge can be expressed as Z 1 1 e2 X ˛ jl .r/C l jti htj˛ jl .r/C l bou ˙ ."a / D 2 .2l C 1/ c d 4 0 "a "t c sgn."t / 0 lD0

(12.87) with a summation over the intermediate bound state jti. Similarly, for the free electron Z 1 1 e2 X ˛ jl .r/C ljq; si hq; sj˛ jl .r/C l free ˙ .!/ D 2 .2l C 1/ c d 4 0 ! "q c sgn."q / 0 lD0

(12.88)

256


summed over free-electron statesjq; ri. Here, ! is the free-running energy parameter and "q represents the energy of the free-electron state jq; si. On the mass shell, p 2 2 ! D "p D c p C m2 c 4 , this becomes ˙ free ."p / D

Z 1 1 l si hq; sj˛ jl .r/C l e2 X ˛ jl .r/C jq; .2l C 1/ c d : 2 4 0 "p "q c sgn."q / 0 lD0

(12.89) The PWR can be combined with the Brown–Langer–Schaefer method discussed above by expanding the remaining terms in (12.83) in a similar way. The free-electron self-energy is diagonal with respect to the momentum p, when all partial waves are included, but this is NOT the case for a truncated sum. Furthermore, it has nondiagonal elements with respect to the spinor index r. 12.3.5.2 Coulomb Gauge The partial-wave regularization has not yet been applied in the Coulomb gauge, but to be able to include the self-energy in a many-body calculation this will be necessary. In analogy with the Feynman-gauge result (12.87), the transverse part of the self-energy in Coulomb gauge becomes Z 1 1 e2 X ˙ bou;trans ."a / D 2 .2l C 1/ c d 4 0 0 lD0

˛jl .r/C ljti htj˛jl .r/C l .˛ r /jl .r/C ljti htj.˛ r /jl .r/C l = 2 "a "t c sgn."t / (12.90) using the expression (4.92). The corresponding mass term becomes in analogy with (12.89) ˙ free;trans ."p / D

Z 1 1 e2 X .2l C 1/ c d 4 2 0 0 lD0

˛jl

.r/C ljq; si

hq; sj˛jl .r/C l .˛ r /jl .r/C ljq; si hq; sj.˛ r /jl .r/C l = 2 : "p "q;s c sgn."q / (12.91)

The Coulomb part in Coulomb gauge is obtained similarly from (4.93) ˙."a /bou;Coul D

1 X e2 sgn." / .2l C 1/ t 8 2 0 r12 lD0 Z 1 2 d jl .r/ C l jti htjjl .r/C l 0

(12.92)

12.4 Dimensional Regularization in Feynman Gauge

257

using the value i sgn."t /=2 for the integral, and the corresponding mass term ˙."a /free;Coul D

1 X e2 sgn." / .2l C 1/ t 8 2 0 r12 lD0 Z 1 2 d jl .r/ C l jq; si hq; sjjl .r/C l :

(12.93)

0

The main advantage of the PWR is that the bound- and free-electron self-energies are calculated in exactly the same way, which improves the numerical accuracy, compared to the standard procedure, where the mass term is evaluated analytically (12.67). Since all terms are here finite, no further regularization is needed. The maximum L value, Lmax , is increased until sufficient convergence is achieved. This scheme has been successfully applied in a number of cases [16, 24]. It has been shown by Persson et al. [22] that the method of PWR gives the correct result in lowest order with an arbitrary number of Coulomb interactions, while a correction term is needed when there is more than one magnetic interaction. This is due to the double summation over partial waves and photon momenta, which is not unique due to the infinities involved. This problem might be remedied by combining this method with dimensional regularization, as will be briefly discussed at the end of the chapter.

12.4 Dimensional Regularization in Feynman Gauge The most versatile regularization procedure developed so far is the dimensional regularization, which is nowadays frequently used in various branches of field theory. In treating the number of dimensions (D) as a continuous variable, it can be shown that the integrals of the radiative effects are singular only when D is an integer. Then by choosing the dimensionality to be 4 , where is a small, positive quantity, the integrals involved will be well defined and finite. After renormalization, one lets the parameter ! 0. This method has been found to preserve the gauge invariance and the validity of the Ward identity to all orders. The method was developed mainly by t’Hooft and Veltman in the 1970s [27] (see, for instance, Mandl and Shaw [17, Chap. 10], Peskin and Schroeder [23, Chap. 7] and Snyderman [25]). Most applications are made in so-called covariant gauges, where the procedure is now well developed. For our purpose, however, it is necessary to apply the Coulomb gauge, and here the procedure is less developed. Important contributions have been made more recently, though, by Adkins [1, 2], Heckarthon [11], and others. Here, we shall first illustrate the method by evaluating the renormalized freeelectron self-energy, using the Feynman gauge. The problem with the Coulomb gauge will be discussed in the next section.

258


12.4.1 Evaluation of the Renormalized Free-Electron Self-Energy in Feynman Gauge We start now from the form (12.12) of the free-electron self-energy in the Feynman gauge and the photon propagator in momentum space (4.28) Z

6 p 6 k C mc d4 k DF .k/ .2/4 .p k/2 m2 c 2 C i

ˇ˙ free .p/ D ie 2 c 2 Z

ie 2 c D 0

6 p 6 k C mc 1 d4 k 2 : 4 2 2 2 .2/ .p k/ m c C i k C i (12.94)

Using the Feynman integral (J.2) (second version) with a D k 2 and b D .p k/2 m2 c 2 , this can be expressed as ˇ˙ free .p/ D

ie 2 c 0

Z

Z

1

dx 0

.6 p 6 k C mc/ d4 k : 4 .2/ Œk 2 C .p 2 2pk m2 c 2 /x2

(12.95)

We shall now evaluate this integral using nonintegral dimension D D 4 , ˇ˙

free

Z

ie 2 c .p/ D 0 2ie 2 c D 0

Z

1

dx 0

Z

Z

1

dx 0

.6 p 6 k C mc/ dD k .2/D Œk 2 C .p 2 2pk m2 c 2 /x2 dD k .1 =2/.6 p 6 k/ .2 =2/ mc ; .2/D Œk 2 C .p 2 2pk m2 c 2 /x2 (12.96)

after applying the anticommutation rules for the gamma matrices in Appendix D.59. With the substitutions q D px and s D .p 2 m2 c 2 /x, this becomes ˇ˙

free

2ie 2 c .p/ D 0

Z

Z

1

dx 0

dD k .1 =2/.6 p 6 k/ .2 =2/ mc ; .2/D Œk 2 C 2kq C s2 (12.97)

which is of the form of (G.23) and (G.24). This leads to ˇ˙

free

2e 2 c.mc/ .p/ D 0 .4/D=2

Z

1

dx 0

. =2/ .2/

Œ.1 =2/.6 p 6 p x/ .2 =2/ mc

m2 c 2 w

=2 (12.98)


259

with w D q 2 s D m2 c 2 p 2 .1 x/ x. The Gamma function can be expanded as (see Appendix G.3) 2 . =2/ D E C with E D 0:5722 : : : being Euler’s constant, and furthermore 1 1 D 1 C ln 4 C ; .4/2 2 .4/D=2 2 2 =2 w m c D 1 ln C : w 2 m2 c 2 This yields . =2/ .4/D=2

m2 c 2 w

=2

1 .2= E C / .4/2

1 C ln 4 C 1 ln w=m2 c 2 C 2 2

w i 1 h ln D C ; (12.99) .4/2 m2 c 2 D

where D

2 E C ln 4 C :

(12.100)

This leads to Z ˇ˙ free .p/ D 2K

h

w i dx .6 p 6 p x 2mc/ ln C 2 2 m c 0 Z 1 dx .6 p 6 p x C mc/ 1

0

(12.101) with KD

e2c c2 ˛ D : 2 0 .4/ 4

We write the denominator in (12.98) as w D m2 c 2 xX I

X D1

with D

p2 .1 x/ D Œ C .1 /x m2 c 2 p 2 m2 c 2 : m2 c 2

(12.102)

260


We then express the integral (12.101) as 2K.A C B C C / with Z AD

1

Z dx .6 p 6 p x 2mc/ C

0

1

dx .6 p 6 p x mc/ ;

0

Z

1

BD

dx .6 p 6 p x 2mc/ ln x;

0

Z

1

C D

dx .6 p 6 p x 2mc/ ln Œ C .1 /x :

0

To evaluate this integral, we can use the formulas (12.65), yielding Z

1

Z dx ln.1 x/ D 1

0

1

dx x ln.1 x/ D 3=4;

0

Z

1

dx ln Œ C .1 /x D 1

0

Z

ln ; 1

ln 1 dx x ln Œ C .1 /x D 1C 1 C 22 ln 2 ; 2 .1 / 1 4.1 /

1 0

which gives A D .6 p =2 2mc/ C .6 p =2 mc/ B D 3 6 p =4 C 2mc 6p ln 6p C D 1C C 1 C 22 ln 2 2 .1 / 1 4.1 / ln .6 p 2m/ 1 C 1 1 1C ln 2 ln D6p C 1C C .1 / 1 2.1 /2 4.1 / ln : C2mc 1 C 1 Subtracting the on-the-mass-shell value (6 p D m; D 0) yields for the A and B terms 6 p mc 1 .A C B/ren D C : 2 2 For the C term, the on-shell value is 5mc=4, yielding C

ren

D 6p

2 ln 3 3 .2 / ln C mc : C C C 2.1 /2 .1 / 4 1 4


261

The total on-shell value (mass-counter term) becomes ımc 2 D

mc 2 ˛ .3 C 4 C / : 4

(12.103)

Collecting all parts, we obtain the following expression for the massrenormalized free-electron self-energy

ˇ˙ free .p/ D

.2 / ln .6 p mc/ C 2 C C 1 .1 /2 2 3 mc 1 ln C 1 1

c2 ˛ 4

(12.104) with D .6 p 2 m2 c 2 /=m2 c 2 . This agrees with the result of Snyderman [25, 26].

12.4.2 Free-Electron Vertex Correction in Feynman Gauge Next, we consider the free-electron vertex correction (12.23) .p 0 ; p/ D

Z

6 p 0 6 k C mc d4 k 0 4 .2/ .p k/2 m2 c 2 C i 6 p 6 k C mc 1 2 : 2 2 2 .p k/ m c C i k C i ie 2 0

(12.105)

The Feynman parametrization (J.4), similar to the self-energy case, a D k 2 , b D .k p/2 m2 c 2 , and c D .k p 0 /2 m2 c 2 , yields .p 0 ; p/ D

2ie 2 0

Z

1

Z

0

Z

1x

d4 k .2/4 0 .6 p 0 6 k C mc/ .6 p 6 k C mc/

dx

dy

3

Œk 2 C .p 2 2pk m2 c 2 /x C .p 02 2p 0 k m2 c 2 /y

:

With q D .px C p 0 y/ and s D p 2 x C p 02 y m2 c 2 .x C y/, the denominator is of the form k 2 C 2kq C s Z Z 1x Z 2ie 2 1 dD k .6 p 0 6 k C mc/ .6 p 6 k C mc/ dx dy 0 0 .2/D .k 2 C 2kq C s/3 0 Z 1x 2 Z 1 2ie D dx dy ŒC0 C C1 C C2 ; c 0 0 0

.p 0 ; p/ D

262


where the index indicates the power of 6 k involved, Z C0 D Z C1 D

dD k .6 p 0 C mc/ .6 p C mc/ ; .2/D .k 2 C 2kq C s/3

dD k . 6 k/ .6 p C mc/ C .6 p 0 C mc/ . 6 k/ ; .2/D .k 2 C 2kq C s/3 Z dD k 6 k 6 k : C2 D D .2/ .k 2 C 2kq C s/3

The coefficients C0 and C1 are convergent and we can let ! 0. With the formula (G.23) (n D 3) and the contraction formulas (D.59), we then have C0 D with

i .6 p C mc/ .6 p 0 C mc/ .4/2 w

w D s q 2 D s .px C p0 y/2 :

Similarly, we have for the numerator in C1 . 6 k/ .6 p Cmc/ C .6 p 0 Cmc/ . 6 k/ D 2.6 p 6 kC 6 k 6 p 0 /8mck and with (G.24) C1 D

i 6 p 6 qC 6 q 6 p 0 4mcq : .4/2 w

The C2 coefficient is divergent and has to be evaluated with more care. Then the situation is analogous to that of the self-energy (12.98). The numerator becomes 6k 6 k 6 k D .2 / 6 k 6 k e 6 ke e and Z C2 D

6 ke e dD k .2 / 6 k 6 k C e 6k ; D 2 3 .2/ .k C 2kq C s/

which can be evaluated with (G.25). The evaluation of the integrals above is straightforward but rather tedious. The complete result is found in [25].

12.5 Dimensional Regularization in Coulomb Gauge

263

12.5 Dimensional Regularization in Coulomb Gauge 12.5.1 Free-Electron Self-Energy in the Coulomb Gauge For our main purpose of combining MBPT and QED, it is necessary to apply the Coulomb gauge to take advantage of the developments in standard MBPT. We shall first follow Adkins [1] in regularizing the free-electron self-energy in the Coulomb gauge by expressing the bound states in terms of free-electron states. We then start from the expression (12.12) Z

6 p 6 k C mc d4 k DF .k/: .2/4 .p k/2 m2 c 2 C i

ˇ˙ free .p/ D ie 2 c 2

(12.106)

For the photon propagator, we use the expressions (4.32) and (4.36) C .kI k/ DF

1 D c 0

ki kj ı;0 ı;0 1 ı;i ı;j gij C 2 : k 2 C i k2 k

(12.107)

The three terms in the propagator correspond to the Coulomb, Gaunt, and scalarretardation parts, respectively, of the interaction (4.59). The Coulomb part of the self-energy becomes Z

1 0 .6 p 6 k C mc/ 0 d4 k ; 2 4 2 2 2 .2/ .p k/ m c C i k C i Z 1 e p e k C mc d4 k ie 2 c D 4 0 .2/ .p k/2 m2 c 2 C i k2 C i ie 2 c 0

(12.108) (12.109)

using the commutation rules in Appendix (D.58). With q D p and s D p 2 m2 c 2 , the denominator is of the form k 2 C 2kq C s and we can apply the formulas (G.26) and (G.27) without any further substitution (n D 1). This gives with k 0 ! q 0 D p 0 , ki ! qi y D pi y, k D i ki ! py, and w D p 2 y 2 C .1 y/yp02 .p 2 m2 c 2 /y Z

1 e p e k C mc dD k 2 D 2 .2/ k C 2kq C s C i k C i Z 1 2 . =2/ e c dy : D p Œp.1 y/ C mc y 0 .4/D=2 0 .w=m2 c 2 /=2

ie 2 c .mc/ 0

Using (12.99), this yields the Coulomb contribution Z

1

K 0

dy p .p .1 y/ C mc/ . ln.yX // y

264


with K D e 2 c=. 0 .4/2 / and w D m2 c 2 y X , X D 1 C .p2 =m2 c 2 /.1 y/. This leads to Z

1

K 0

dy .p .1 y/ C mc ln y ln X p y

and the Coulomb part becomes (times K)

Z 1 dy 4 32 p C 2mc C p C 4mc p ..p .1 y/ C mc/ ln X: 3 9 y 0 (12.110)

The Gaunt term becomes, using (12.106) and the second term of (12.107),

ie 2 c 0

Z

1 i .6 p 6 k C mc/ i d4 k : 4 .2/ .p k/2 m2 c 2 C i k 2 C i

(12.111)

The denominator is here the same as in the treatment of the self-energy in the Feynman gauge, and we can use much of the results obtained there.4 In analogy with (12.95), we then have

Z

Z

i .6 p 6 k C mc/ i d4 k 4 .2/ Œk 2 C .p 2 2pk m2 c 2 /x2 0 Z Z ie 2 c 1 d4 k .3 /mc .2 /.6 p 6 k/ e p Ce k D dx ; (12.112) 2 4 2 2 2 2 0 0 .2/ Œk C .p 2pk m c /x

ie 2 c 0

1

dx

after inserting the Feynman integral (J.2) and applying the commutation rules in Appendix (D.59). With the substitutions k ! q D px and s D .p 2 m2 c 2 /x, the equation above leads after applying (G.23) and (G.24) in analogy with (12.98) to

ie 2 c 0

Z

Z

1

dx 0

e2 c 0 .4/D=2 e2 c D 0 .4/D=2 D

dD k .3 /mc .2 /.6 p 6 k/ e p Ce k 2 D 2 .2/ Œk C 2kq C s/ Z 1 . =2/ dx Œ.3 /mc .2 / 6 p .1 x/ e p .1 x/ .w= /=2 0 Z 1 dx .1 x/ 3 0 p0 p C 3mc 0

C ..1 x/ 6 p mc/

4

. =2/ ; .w= /=2

We use the convention that ; ; : : : represent all four components (0,1,2,3), while i; j; : : : represent the vector part (1,2,3).

12.5 Dimensional Regularization in Coulomb Gauge

265

where w D q 2 s D p 2 x 2 .p 2 m2 c 2 /x D m2 c 2 xY . This yields (times K) Z

1

dx 0

˚ .1 x/ 3 0 p0 p 3mc Œ ln.xY / 2 ..1 x/ 6 p mc/g

using the relation (12.99) and the fact that ! 2 as ! 0. Then the Gaunt part becomes 1 0 5 1 3 p0 p 3mc 0 p0 p C mc 2 4 4 Z 1 0 C dx .1 x/ 3 p0 p 3mc ln Y:

(12.113)

0

Finally, the scalar-retardation part becomes similarly, using the third term of (12.107) and the commutation rules (D.57),

ie 2 c 0 D

Z

1 d4 k i ki .6 p 6 k C mc/ j kj 1 2 4 2 2 2 2 .2/ .p k/ m c C i k k C i

ie 2 c 0

Z

ie 2 c D 0

1 d4 k i ki j kj .6 p 6 kmc/ 2 i ki .k j pj k j kj / 1 .2/4 .pk/2 m2 c 2 C i k2 k 2 C i Z

1 k mc C 2 i ki k j pj =k2 d4 k 6 p e .2/4 .p k/2 m2 c 2 C i k 2 C i

with i ki j kj D k2 D ki ki . With the same substitutions as in the Gaunt case, this becomes

ie 2 c 0

Z

Z

1

dx 0

k mc C 2 i ki k j pj =k2 dD k 6 p e : D .2/ Œk 2 2pkx C .p 2 m2 c 2 /x2

(12.114)

With the substitutions k ! q D px and s D .p 2 m2 /x the first part is of the form (G.23) and (G.24) and becomes e2c 0 .4/D=2

Z 0

1

dx Œ6 p e p x mc

. =2/ w=2

(12.115)

and with (12.99) Z K 0

1

dx Œ6 p e p x mc . ln.xY //

(12.116)

266


with K D e 2 c=. 0 .4/2 / and w being the same as in the Gaunt case, w D q 2 s D p 2 x 2 p 2 x C m2 c 2 x D m2 c 2 xY . The second part of (12.114) is of the form (G.28) and becomes (ki k j ! qi q j 2 y D pi p j x 2 y 2 in first term, ! 12 gji D 12 ıij in second)

.1 C =2/ . =2/ j p ; j w1C=2 w=2 0 0 i Z 1 Z 1 2 pi p j pj xy p j dx dy y pj . ln.xyZ// ; K m2 c 2 Z 0 0 Z 1 Z 1 2p p2 xy p dx dy y C p . ln.xyZ// DK m2 c 2 Z 0 0 Z

Z

1

K

1

dx

dy

p

y

2 i pi p i pj p j

with w D xy p2 xy C p02 x p 2 C m2 c 2 D xyŒp2 .1 xy/ p02 .1 x/ C m2 c 2 D m2 c 2 Zxy. Integration by parts of the first term yields (times K), noting that dZ=dy D p2 x, Z

1

dx 0

p

y y 2p ln Z

Z

1

C3 0

Z

1

1

dx 0

dy

p y p ln Z:

0

The total scalar-retardation part then becomes (with Z.y D 1/ D Y ) Z

1 0

Z 1 dx Œ6 p e p x mc . ln.xY // dx 2p ln Y 0 Z 1 Z 1 Z 1 Z 1 p p C3 dx dy y p ln Z C dx dy y p . ln.xyZ// 0

0

0

0

or Z

Z

1 p dx p0 .1 x/ p.1 C x/ mc . ln x/ C dy y p 0 0 Z 1 0 dx p0 .1 x/ p.1 x/ mc ln Y 0 Z 1 Z 1 Z 1 Z 1 p p dx dy y p ln.xy/ C 2 dx dy y p ln.xy/ 0 0 0 0 Z 1 Z 1 p 3 dx dy y p ln Z; 1

0

0

0

12.6 Direct Numerical Regularization of the Bound-State Self-Energy

267

which gives5

5 5 3 1 0 p0 p mc C 0 p0 p mc 2 6 4 36 Z 1 dx 0 p0 .1 x/ p.1 x/ mc ln Y 0 Z 1 Z 1 p dx dy y p ln Z: 3 0

0

Summarizing all contributions yields the mass-renormalized free-electron selfenergy in Coulomb gauge " Z 1 1 0 19 e2c dy .6 p mc/ p C p p .p .1 y/ C mc/ ln X 0 2 6 y 0 .4/2 0 # Z 1 Z 1 Z 1 p dx Œ.1 x/ 6 p mc ln Y C dx dy y 2p ln Z ; C2 0

0

0

(12.117) where we have subtracted the on-shell (6 p D mc) value, mc.3 C 4/. (The expressions for X; Y; Z are given in the text.) This is in agreement with the the result of Adkins [1]. The treatment of the vertex correction is more complex and will not be reproduced here. Interested readers are referred to the papers by Adkins.6

12.6 Direct Numerical Regularization of the Bound-State Self-Energy As an alternative to the regularization procedure discussed above, we shall consider a new more direct procedure, where the regularization is performed directly in the bound state, without any transformation to free-electron states. This is presently being tested by the Gothenburg group [12].

5

Z

Z

1

dx 0

6

1

dy 0

Z 1 Z 1 p p 10 1 y ln.xy/ D I dx x dy y ln.xy/ D I 9 18 0 0 Z 1 Z 1 p 9 dx x dy y y ln.xy/ D : 50 0 0

A complete treatment of the vertex correction is being published separately in arXiv:quant-ph.

268


12.6.1 Feynman Gauge The bound-state self-energy in the Feynman gauge is from (8.48) ˇZ ˇ hrj˙."a /jai D rt ˇˇ

ˇ

ˇ c d f F ./ ˇ ta ; "a "t .c i/t ˇ

(12.118)

where D jkj and the function f F is given by (4.55). The integral over is convergent, while the summation over t is (logarithmically) divergent. With 3 dimensions of the k-vector space, we make the substitution Z

d3 k ) .2/3

Z

Z d˝

2 d .2/3

and the self-energy becomes ˇ ˇ

Z ˇ ˛1 ˛2 c d sin r12 ˇˇ e 2 .2/ ˇ at ta; haj˙ ."a /jai D ˇ r 4 2 0 "a "t .c i/t ˇ 12

(12.119)

which would make the expression convergent for ¤ 0. In a similar way, the the free-electron self-energy can be expressed. In analogy with the expression (12.85), this leads to the renormalized bound-state self-energy ˇ ˛ ˇ ˛ ˝ ˇ bou ˝ ˇ r ˇ˙ren ."a /ˇa D lim r ˇ˙bou ."a / ˙free ."a /ˇa !0 ˇ ˛ ˇ ˛ ˇ ˝ ˝ ˇ C r ˇp0 ; r 0 i p0 ; r 0 ˇ˙free ."a / ˙free ."p /ˇp; r hp; r ˇa : (12.120)

12.6.2 Coulomb Gauge The transverse part of the self-energy expression in Coulomb gauge is in analogy with the Feynman expression (12.118) ˇZ ˇ c d fTC ./ D rt ˇˇ "a "t .c i/t

hrj˙."a /jaiTrans

ˇ

ˇ ˇ ta ; ˇ

(12.121)

where fTC is given by (4.60). The Coulomb part is given by (8.49) hbj˙."a /jaiCoul D

ˇ

ˇ Z ˇ 2 d sin r12 ˇˇ 1 e2 sgn."t / bt ˇˇ 2 ˇ ta (12.122) 2 4 0 r12 2

and the renormalization can be performed as in the Feynman gauge. The procedure of direct numerical regularization outlined here is presently being tested by the Gothenburg group.

References

269

References 1. Adkins, G.: One-loop renormalization of Coulomb-gauge QED. Phys. Rev. D 27, 1814–20 (1983) 2. Adkins, G.: One-loop vertex function in Coulomb-gauge QED. Phys. Rev. D 34, 2489–92 (1986) 3. Bethe, H.A.: The Electromagnetic Shift of Energy Levels. Phys. Rev. 72, 339–41 (1947) 4. Bjorken, J.D., Drell, S.D.: Relativistic Quantum Mechanics. Mc-Graw-Hill Pbl. Co, N.Y. (1964) 5. Blundell, S., Snyderman, N.J.: Basis-set approach to calculating the radiative self-energy in highly ionized atoms. Phys. Rev. A 44, R1427–30 (1991) 6. Brown, G.E., Langer, J.S., Schaeffer, G.W.: Lamb shift of a tightly bound electron. I. Method. Proc. R. Soc. London, Ser. A 251, 92–104 (1959) 7. Brown, G.E., Mayers, D.F.: Lamb shift of a tightly bound electron. II. Calculation for the K-electron in Hg. Proc. R. Soc. London, Ser. A 251, 105–109 (1959) 8. Cheng, K.T., Johnson, W.R.: Self-energy corrections to the K-electron binding energy in heavy and superheavy atoms. Phys. Rev. A 1, 1943–48 (1976) 9. Cheng, T.K., Johnson, W.R., Sapirstein, J.: Screend Lamb-shift calculations for Lithiumlike Uranium, Sodiumlike Platimun, and Copperlike Gold. Phys. Rev. Lett. 23, 2960–63 (1991) 10. Desiderio, A.M., Johnson, W.R.: Lamb Shift and Binding Energies of K Electrons in Heavy Atoms. Phys. Rev. A 3, 1267–75 (1971) 11. Heckarthon, D.: Dimensional regularization and renormalization of Coulomb gauge quantum electrodynamics. Nucl. Phys. B 156, 328–46 (1978) 12. Hedendahl, D.: p. (private communication) 13. Itzykson, C., Zuber, J.B.: Quantum Field Theory. McGraw-Hill (1980) 14. Labzowsky, L.N., Mitrushenkov, A.O.: Renormalization of the second-order electron selfenergy for a tightly bound atomic electron. Phys. Lett. A 198, 333–40 (1995) 15. Lindgren, I., Persson, H., Salomonson, S., Sunnergren, P.: Analysis of the electron self-energy for tightly bound electrons. Phys. Rev. A 58, 1001–15 (1998) 16. Lindgren, I., Persson, H., Salomonson, S., Ynnerman, A.: Bound-state self-energy calculation using partial-wave renormalization. Phys. Rev. A 47, R4555–58 (1993) 17. Mandl, F., Shaw, G.: Quantum Field Theory. John Wiley and Sons, New York (1986) 18. Mohr, P.J.: Numerical Evaluation of the 1s1=2 -State Radiative Level Shift. Ann. Phys. (N.Y.) 88, 52–87 (1974) 19. Mohr, P.J.: Self-Energy Radiative Corrections. Ann. Phys. (N.Y.) 88, 26–51 (1974) 20. Mohr, P.J.: Self-energy of the n D 2 states in a strong Coulomb field. Phys. Rev. A 26, 2338–54 (1982) 21. Pauli, W., Willars, F.: On the Invariant Regularization in Relativistic Quantum Theory. Rev. Mod. Phys. 21, 434–44 (1949) 22. Persson, H., Salomonson, S., Sunnergren, P.: Regularization Corrections to the Partial-Wave Renormalization Procedure. Presented at the conference Modern Trends in Atomic Physics, Hind˚as, Sweden, May, 1996. Adv. Quantum Chem. 30, 379–92 (1998) 23. Peskin, M.E., Schroeder, D.V.: An introduction to Quantun Field Theory. Addison-Wesley Publ. Co., Reading, Mass. (1995) 24. Quiney, H.M., Grant, I.P.: Atomic self-energy calculations using partial-wave mass renormalization. J. Phys. B 49, L299–304 (1994) 25. Snyderman, N.J.: Electron Radiative Self-Energy of Highly Stripped Heavy-Atoms. Ann. Phys. (N.Y.) 211, 43–86 (1991) 26. Sunnergren, P.: Complete One-Loop QED CAlculations for Few-Eleectron Ions. Ph.D. thesis, Department of Physics, Chalmers University of Technology and University of Gothenburg, Gothenburg, Sweden (1998) 27. t’Hooft, G., Veltman, M.: Regularization and renormalization of gauge fields. Nucl. Phys. B 44, 189–213 (1972)

Chapter 13

Summary and Conclusions

The all-order forms of many-body perturbation theory (MBPT), like the coupled-cluster approach (CCA), have been extremely successful in calculations on atomic and in particular on molecular systems. Here, the dominating parts of the electron correlation can be evaluated to essentially all orders of perturbation theory. A shortcoming, however, of the standard MBPT/CCA procedures is that quantum-electrodynamics (QED) can only be included in a very limited fashion (first-order energy). Particularly for highly charged systems, QED effects can be quite important. Certain experimental data on such systems are now several orders of magnitude more accurate than the best available theoretical calculation. It is believed that this shortcoming is due to the omitted combination of many-body and QED effects in presently available theoretical procedures. The procedure presented in this book, which is based on quantum-field theory, describes – for the first time – a road toward a rigorous unification of QED and MBPT. The procedure is based on the covariant evolution operator, which describes the time evolution of the relativistic wave function or state vector. The procedure is for two-electron systems fully compatible with the relativistically covariant Bethe– Salpeter equation, but it is more versatile. The procedure is – in contrast to the standard Bethe–Salpeter equation – applicable to a general multi-reference (quasi-degenerate) model space. It can also be combined with the coupled-cluster approach and is, in principle, applicable to systems with more than two electrons. The covariant evolution operator contains generally singularities that can be eliminated. The regular part is referred to as the Green’s operator, which is a generalization of the Green’s-function concept. In principle, the Green’s operator – as well as the Bethe–Salpeter equation – has individual times for the particles involved. Most applications, though, use the equaltime approximation, where the times are equalized, to make the procedure consistent with the quantum-mechanical picture. The Green’s operator for time t D 0 corresponds to the wave operator used in standard MBPT, and the time derivative at t D 0, operating within the model space, to the many-body effective interaction. This connects the field-theoretical procedure with the standard MBPT.


271

272

13 Summary and Conclusions

The formalism presented here has been partially tested numerically by the Gothenburg atomic theory group, and in cases where comparison can be made with the more restricted S-matrix formulation, very good agreement is reported. A big challenge is the renormalization of the radiative effects, which generally has to be performed using the Coulomb gauge, to take advantage of the developments in MBPT/CCA. Schemes have been developed for this process but so far not been implemented in a QED-MBPT procedure. When the procedure is more developed, critical tests can be performed to find out to what extent the new effects will improve the agreement between theory and accurate experimental data.

Appendix A

Notations and Definitions

A.1 Four-Component Vector Notations A four-dimensional contravariant vector is defined as1 x D x D .x 0 ; x 1 ; x 2 ; x 3 / D .x 0 ; x/ D .ct; x/;

(A.1)

where D0; 1; 2; 3 and x is the three-dimensional coordinate vector xD.x 1 ; x 2 ; x 3 / .x; y; z/. The four-dimensional differential is d 4 x D cdt d3 x

and d3 x D dx dy dz:

A corresponding covariant vector is defined as x D .x0 ; x1 ; x2 ; x3 / D g x D .x0 ; x/ D .ct; x/;

(A.2)

which implies that x 0 D x0

x i D xi

.i D 1; 2; 3/;

(A.3)

g is a metric tensor, which can raise the so-called Lorentz indices of the vector. Similarly, an analogous tensor can lower the indices x D g x :

(A.4)

These relations hold generally for four-dimensional vectors. There are various possible choices of the metric tensors, but we shall use the following 0 1 1 0 0 0 B 0 1 0 0 C C g D g D B (A.5) @ 0 0 1 0 A : 0 0 0 1

1

In all appendices, we display complete formulas with all fundamental constants. As before, we use the Einstein summation rule with summation over repeated indices.

I. Lindgren, Relativistic Many-Body Theory: A New Field-Theoretical Approach, Springer Series on Atomic, Optical, and Plasma Physics 63, c Springer Science+Business Media, LLC 2011 DOI 10.1007/978-1-4419-8309-1,

273

274

A Notations and Definitions

The four-dimensional scalar product is defined as the product of a contravariant and a covariant vector: ab D a b D a0 b0 a b;

(A.6)

where a b is the three-dimensional scalar product a b D ax bx C ay by C az bz : The covariant gradient operator is defined as the gradient with respect to a contravariant coordinate vector: @ 1 @ ;r (A.7) @ D D @x c @t and the contravariant gradient operator analogously @ @ D D @x

1 @ ; r ; c @t

(A.8)

r is the three-dimensional gradient operator rD

@ @ @ eO x C eO y C eO z ; @x @y @z

where .eO x ; eO y ; eO z / are unit vectors in the coordinate directions. The four-dimensional divergence is defined as @ A D

1 @A0 C r A D rA; c @t

(A.9)

where r A is the three-dimensional divergence r A D

@Ax @Ay @Az C C : @x @y @z

The d’Alembertian operator is defined as D @ @ D where r2 D D is the Laplacian operator.

1 @2 r 2 D r2; c 2 @t 2 @2 @2 @2 C C @x 2 @y 2 @z2

(A.10)

A.2 Vector Spaces

275

A.2 Vector Spaces A.2.1 Notations X; Y; : : : are sets with elements x; y; : : :. x 2 X means that x is an element in the set X . N is the set of nonnegative integers. R is the set of real numbers. C is the set of complex numbers. R n is the set of real n-dimensional vectors. Cn is the set of complex n-dimensional vectors. A X means that A is a subset of X . A [ B is the union of A and B. A \ B is the intersection of A and B. A D fx 2 X W P g means that A is the set of all elements x in X that satisfy the condition P . f W X ! Y represents a function or operator, which means that f maps uniquely the elements of X onto elements of Y . A functional is a unique mapping f W X ! R .C/ of a function space on the space of real (complex) numbers. The set of arguments x 2 A for which the function f W A ! B is defined is the domain, and the set of results y 2 B which can be produced is the range. .a; b/ is the open interval fx 2 R W a < x < bg. Œa; b is the closed interval fx 2 R W a x bg. sup represents the supremum, the least upper bound of a set inf represents the infimum, the largest lower bound of a set.

A.2.2 Basic Definitions A real (complex) vector space or function space X is an infinite set of elements, x, referred to as points or vectors, which is closed under addition, x C y D z 2 X , and under multiplication by a real (complex) number c, cx D y 2 X . The continuous functions f .x/ on the interval x 2 Œa; b form a vector space, also with some boundary conditions, like f .a/ D f .b/ D 0. A subset of X is a subspace of X if it fulfills the criteria for a vector space. A norm of a vector space X is a function p W X ! Œ0; 1 with the properties (1) p.x/ D jjp.x/ (2) p.x C y/ p.x/ C p.y/ for all real ( 2 R) and all x; y 2 X (3) that p.x/ D 0 always implies x D 0

The norm is written as p.x/ D jjxjj. We then have jjxjj D jj jjxjj , jjx C yjj jjxjj C jjyjj , and jjxjj D 0 ) x D 0. If the last condition is not fulfilled, it is a seminorm.

276


A vector space with a norm for all its elements is a normed space, denoted by .X; jjjj/. The continuous functions, f .x/, on the interval Œa; b form a normed Rb space by defining a norm, for instance, jjf jj D Œ a dt jf .t/j2 1=2 . By means of the Cauchy–Schwartz inequality, it can be shown that this satisfies the criteria for a norm [4, p. 93]. If f is a function f W A ! Y and A X , then f is defined in the neighborhood of x0 2 X , if there is an > 0 such that the entire sphere fx 2 X W jjx x0 jj < g belongs to A [4, p. 309]. A function/operator f W X ! Y is bounded, if there exists a number C such that jjf xjj D C < 1: sup jjxjj 0¤x2X Then C D jjf jj is the norm of f . Thus, jjf xjj jjf jj jjxjj . A function f is continuous at the point x0 2 X , if for every ı > 0 there exists an > 0 such that for every member of the set x W jjx x0 jj < we have jjf x f x0 jj ı [4, p. 139]. This can also be expressed so that f is continuous at the point x0 , if and only if f x ! f x0 whenever xn ! x0 , fxn g being a sequence in X , meaning that f xn converges to f x0 , if x converges to x0 [5, p. 70]. A linear function/operator is continuous if and only if it is bounded [4, p. 197, 213], [2, p. 22]. A functional f W X ! R is convex if f .tx C .t 1/y/ tf .x/ C .t 1/f .y/; for all x; y 2 X and t 2 .0; 1/. A subset A X is open, if for every x 2 A there exists an > 0 such that the entire ball Br .x/ D fy 2 X j jjy xjj < g belongs to A, i.e., Br .x/ A [1, p. 363], [4, p. 98], [5, p. 57]. A sequence fxn g , where n is an integer (n 2 N), is an infinite numbered list of elements in a set or a space. A subsequence is a sequence, which is a part of a sequence. A sequence fxn 2 Ag is (strongly) convergent toward x 2 A, if and only if for every > 0 there exists an N such that jjxn xjj < for all n > N [4, p. 95, 348]. A sequence is called a Cauchy sequence if and only if for every > 0 there exists an N such that jjxn xm jj < for all m; n > N . If a sequence fxn g is convergent, then it follows that for n; m > N jjxm xn jj D jj.xn x/ C .x xm /jj jjxn xjj C jjxm xjj < 2; which means that a convergent sequence is always a Cauchy sequence. The opposite is not necessarily true, since the point of convergence need not be an element of X [3, p. 44].

A.3 Special Functions

277

A subset A of a normed space is termed compact, if every infinite sequence of elements in A has a subsequence, which converges to an element in A. The closed interval [0,1] is an example of a compact set, while the open interval (0,1) is noncompact, since the sequence 1, 1/2, 1/3. . . and all of its subsequences converge to 0, which lies outside the set [5, p. 149]. This sequence satisfies the Cauchy convergence criteria but not the (strong) convergence criteria. A dual space or adjoint space of a vector space X , denoted as X , is the space of all functions on X . An inner or scalar product in a vector space X is a function h; i W X X ! R with the properties (1) hx; 1 y1 C 2 y2 i D 1 hx; y1 i C 2 hx; y2 i ;

hx; yi D hy; xi;

for all x; y; y1 ; y2 2 X and all j 2 R , and (2) hx; xi D 0 only if x D 0.

A.2.3 Special Spaces A.2.3.1 Banach Space A Banach space is a normed space in which every Cauchy sequence converges to a point in the space.

A.2.3.2 Hilbert Space p A Banach space with the norm jjxjj D C hx; xi is called a Hilbert space [1, p. 364]. A.2.3.3 Fock Space A Fock space is a Hilbert space, where the number of particles is variable or unknown.

A.3 Special Functions A.3.1 Dirac Delta Function We consider the integral

Z

L=2 L=2

dx eikx :

(A.11)

278


Assuming periodic boundary conditions, eiLx=2 D eiLx=2 , limits the possible k values to k D kn D 2 n=L. Then 1 L

Z

L=2

L=2

dx eikn x D ıkn ;0 D ı.kn ; 0/;

(A.12)

where ın;m is the Kronecker delta factor ın;m D

1 if m D n : 0 if m ¤ n

(A.13)

If we let L ! 1, then we have to add a “damping factor” e jxj , where is a small positive number, to make the integral meaningful, Z

1

dx eikx e jxj D

1

k2

2 D 2 .k/: C 2

(A.14)

dx eikx e jxj D ı.k/;

(A.15)

In the limit ! 0, we have 1 lim 2 !0

lim .k/ D

!0

Z

1 1

which can be regarded as a definition of the Dirac delta function, ı.k/. Formally, we write this relation as Z 1 dx ikx (A.16) e D ı.k/: 2 1 The function also has the following properties Z

lim .x/ D ıx;0 ;

1 1

!0

dx .x a/ .x b/ D C .a b/:

(A.17)

In three dimensions, (A.12) goes over into 1 V

Z V

d3 x eikn x D ı 3 .kn ; 0/ D ı.knx ; 0/ ı.kny ; 0/ ı.knz ; 0/:

(A.18)

In the limit where the integration is extended over the entire three-dimensional space, we have in analogy with (A.16) Z

d3 x ikx e D ı 3 .k/: .2/3

(A.19)


279

A.3.2 Integrals over Functions We consider the integral Z

Z

1

1

dx ı.x a/ f .x/ D lim

dx .x a/ f .x/ Z 1 1 2 D lim dx f .x/: (A.20) 2 !0 1 .x a/2 C 2 !0 1

1

The integral can be evaluated using residue calculus and leads to Z

1

dx ı.x a/ f .x/ D f .a/

(A.21a)

1

provided the function f .x/ has no poles. In three dimensions, we have similarly Z d3 x ı.x x 0 / f .x/ D f .x 0 /

(A.21b)

integrated over all space. The relations above are often taken as the definition of the Dirac delta function, but the procedure applied here is more rigorous. Next, we consider the integral over two functions Z 2 2 1 dx .2/2 .x a/2 C 2 .x b/2 C 2 Z 1 1 1 D dx 4 2 i x a i x a C i 2 .x b C i /.x b i / 1 1 1 D 2i b a i. C / b a C i. C / 1 2. C / D (A.22) 2 .a b/2 C . C /2

Z

dx .x a/ .x b/ D

after integrating the first term over the negative and the second term over the positive half plane. Thus, Z dx .x a/ .x b/ D C .a b/ and we see that here the widths of the functions are added.

(A.23)

280


Now, we consider some integrals with the functions in combination with electron and photon propagators that are frequently used in the main text. First, we consider the integral with one function and an electron propagator (4.10) Z d!

1 ."a !/ D ! "j C i

Z

d! 1 2 2 ! "j C i ."a !/2 C 2 Z d! 1 2 D : 2 ! "j C i ."a ! C i /."a ! i /

The pole of the propagator yields the contribution ."a "j /, which vanishes in the limit ! 0, if "a 6D "j . Nevertheless, we shall see that this pole has a significant effect on the result. Integrating above over the positive half plane, with the single pole "a C i , yields 1 "a "j C i C i and integrating over the negative half plane, with the two poles "j i ; "a i , yields

1 2i C ."a "j C i C i /."a "j i C i / "a "j i C i 1 ; D "a "j C i C i

which is identical to the previous result. We observe here that the pole of the propagator, which has a vanishing contribution in the limit ! 0, has the effect of reversing the sign of the i term. The parameter originates from the adiabatic damping and is small but finite, while the parameter is infinitely small and only determines the position of the pole of the propagator. Therefore, if they appear together, the term dominates, and the term can be omitted. This yields Z d!

1 1 ."a !/ D ! "j C i "a "j C i

(A.24)

noting that the parameter of the propagator is replaced by the damping parameter . Second, we consider the integral with the photon propagator (4.31) Z Z 1 1 d! 1 1 d! 2 ." !/ D a ! 2 C i 2

2 ! C i ! C i 2 ."a ! C i /."a ! i /


281

D D

1 1 1 2 "a C i C i "a i C i "2a

1 . i i /2

(A.25)

or, neglecting the term, Z d!

!2

1 1 ."a !/ D 2 2 C i "a 2 C i

(A.26)

noting that 0. Finally, we consider the integrals of two functions and the propagators. With the electron propagator, we have Z

1 ."a !/ ."b !/ ! "j C i Z 1 1 1 1 d! D .2i/2 ! "j C i "a ! i "a ! C i 1 1 : "b ! i "b ! C i

d!

Here, three of the combinations with poles on both sides of the real axis contribute, which yields 1 1 1 C 2i ."b "j C i /."a "b 2i / ."a "j C i /."b "a 2i / 1 : ."a "j C i /."b "j C i / The last two terms become 1 1 1 1 "a "j C i "b "a 2i "b "j C i ."b "a 2i /."b "j C i / neglecting an imaginary term in the numerator. This leads to Z d!

1 1 ."a !/ ."b !/ 2 ."a "b /: ! "j C i "a "j C i (A.27)

282


Similarly, we find for the photon propagator Z d!

1 1 ."a !/ ."b !/ 2 ."a "b /: ! 2 2 C i "a 2 C i (A.28)

Formally, we can obtain the integral with propagators by replacing the function by the corresponding Dirac delta function, noting that we then have to replace the imaginary parameter in the denominator by the damping factor .

A.3.3 The Heaviside Step Function The Heaviside step function is defined as .t/ D 1 D0

t0 > t t 0 < t:

(A.29)

The step function can also be given the integral representation Z

1

.t/ D lim i !0

1

d! ei!t 2 ! C i

(A.30)

from which we obtain the derivative of the step function d.t/ D lim !0 dt

Z

1 1

d! ! ei!t D ı.t/ 2 ! C i

(A.31)

where ı.t/ is the Dirac delta function.

References 1. Blanchard, P., Brüning, E.: Variational Methods in Mathematical Physics. A Unified Approach. Springer-Verlag, Berlin (1992) 2. Debnath, L., Mikusiński, P.: Introduction to Hilbert Spaces with Applications., second edn. Academic Press, New York (1999) 3. Delves, L.M., Welsh, J.: Numerical Solution of Integral Equations. Clarendon Press, Oxford (1974) 4. Griffel, D.H.: Applied Functional Analysis. Wiley, N.Y. (1981) 5. Taylor, A.E.: Introduction to Functional Analysis. Wiley and Sons (1957)

Appendix B

Second Quantization

B.1 Definitions (See, for instance [2, Chap. 5], [1, Chap. 11].) In second quantization – also known as the number representation – a state is represented by a vector (see Appendix C.1) jn1 ; n2 ; i, where the numbers represent the number of particles in the particular basis state (which for fermions can be equal only to one or zero). Second quantization is based on annihilation/creation operators cj /cj , which annihilate and create, respectively, a single particle. If we denote byj0i the vacuum state with no particle, then cjj0i Djj i (B.1) represents a single-particle state. In the coordinate representation (C.19), this corresponds to the wave function j .x/ D hxjj i (B.2) satisfying the single-electron Schrödinger or Dirac equation. Obviously, we have cj j0i D 0:

(B.3)

For fermions, the operators satisfy the anticommutation relations

fci ; cj g D ci cj C cj ci D 0; fci ; cj g D ci cj C cj ci D 0; fci ; cj g D ci cj C cj ci D ıij ;

(B.4)

where ıij is the Kronecker delta factor (A.13). It then follows that ci cjj0i D cj cij0i ;

(B.5)

283

284

B Second Quantization

which means that ci ci j0i represents an antisymmetric two-particle state, which we denote in the following way1 ci cjj0i Djfi; j gi :

(B.6)

The antisymmetric form is required for fermions by the quantum-mechanical rules. A corresponding bra state is h0jcl ck D hfk; lgj

(B.7)

and it then follows that the states are orthonormal. In the coordinate representation, the state above becomes 1 hx 1 x 2jfi; j gi D p i .x 1 / j .x 2 / j .x 1 / i .x 2 / : 2

(B.8)

Generalizing this to a general many-particle system leads to an antisymmetric product, known as the Slater determinant, 1 Detfa; b; N g hx 1 ; x 2 ; x N jca cb cN j0i D p NŠ ˇ ˇ ˇ 1 .x 1 / 1 .x 2 / 1 .x N / ˇ ˇ ˇ 1 ˇˇ 2 .x 1 / 2 .x 2 / 2 .x N / ˇˇ D p ˇ ˇ : (B.9) NŠ ˇ ˇ ˇ .x / .x / .x / ˇ N 1 N 2 N N For an N -particle system, we define one- and two-particle operators by F D GD

n X

fn ;

nD1 n X

gmn ;

(B.10)

(B.11)

m t1

: O Hint;I .t1 / HO int;I .t2 /

t1 > t2

:

(F.57)

In the Coulomb gauge, the interaction is given by (F.36) and the vector potential is given by (F.39). The evolution operator for the combined interactions will then be U.2/ .0; 1/ D cr ca cs cb e

e2 c2 „2

Z

Z

0

0

dt1 T 1 1 it1 ."a "r Ci /=„ it2 ."b "s Ci /=„

Here T Œ.˛ A ? /1 .˛ A ? /2 D

dt2

e

X kp

Œ.˛ A ? /1 .˛ A ? /2

:

(F.58)

„ .˛ "p /1 .˛ "p /2 2!0 V

8 i.k x !t / i.k x !t / < e 1 1 1 e 2 2 2 t2 > t1 :

ei.k2 x2 !t2 / ei.k1 x1 !t1 / t1 > t2

or with r 12 D x 1 x 2 and t12 D t1 t2 2 1 X ei.kr 12 !t12 / „ X T Œ.˛ A ? /1 .˛ A ? /2 D .˛ "p /1 .˛ "p /2 ; 0 pD1 V k 2!

(F.59)

322

F Semiclassical Theory of Radiation

where the upper sign is valid for t2 > t1 . This yields U.2/ .0; 1/

D

cr ca cs cb

e2 c2 0 „2

Z

Z

0

1

dt2

0

1

dt1

2 X

.˛ "p /1 .˛ "p /2

pD1

1 X ei.kr 12 !t12 / icq t12 .t1 Ct2 / e e V k 2!

(F.60)

using the energy conservation (F.49). The boxed part of the equation above is essentially the photon propagator (4.23) Z d3 k ei.kr 12 !t12 / 1 X ei.kr 12 !t12 / ) : DF .1; 2/ D V 2! .2/3 2!

(F.61)

k

This can be represented by a complex integral Z Z dz d3 k eizt12 DF .1; 2/ D i eikr 12 ; 3 2 .2/ 2 z ! 2 C i

(F.62)

where is a small, positive quantity. As before, the sign of the exponent ik r 12 is immaterial. The integrand has poles at z D ˙.! i /, assuming ! to be positive. 1 i! t12 ikr 12 For t2 > t1 , integration over the negative half plane yields 2! e e and for 1 i! t12 ikr 12 e , which is t1 > t2 integration over the positive half plane yields 2! e identical to (F.61). The evolution operator (F.60) can then be expressed as Z 0 Z e2 c2 0 dt2 dt1 U.2/ .0; 1/ D cr ca cs cb 0 „ 1 1 2 X .˛ "p /1 .˛ "p /2 DF .1; 2/ eicq t12 =„ e.t1 Ct2 /„ : pD1

(F.63)

F.2.4 Comparison with the Covariant Treatment It is illuminating to compare the quantization of the transverse photons with the fully covariant treatment, to be discussed in the next chapter. Then we simply have to replace the sum in (F.42) by the corresponding covariant expression 2 X p1 p2 D1

.akp ˛ "p /1 .akp ˛ "p /2 )

3 X

.akp ˛ "p /1 .akp ˛ "p /2 :

p1 p2 D0

(F.64)

F.2 Quantized Radiation Field

323

The commutation relation (G.10) yields 3 X

.akp ˛ "p /1 .akp ˛ "p /2 D ˛1 ˛2 1:

(F.65)

p1 p2 D0

We then find that the equivalent potential interaction (F.50) under energy conservation is replaced by V12

e2 D 0

Z

d3 k ei kr 12 .1 ˛ ˛ / 1 2 .2/3 q 2 k 2 C i

(F.66)

and with the Fourier transform given in Appendix J.2 V12 D

e2 .1 ˛1 ˛2 / eijqjr12 : 40 r12

(F.67)

We shall now compare this with the exchange of transverse photons, treated above. We then make the decomposition

1 ˛1 ˛ 2 D

8 O O < 1 .˛1 k/.˛ 2 k/ :

O O ˛1 ˛2 C .˛1 k/.˛ 2 k/

:

(F.68)

The last part, which represents the exchange of transverse photons, is identical to (F.52), which led to the Breit interaction. The first part, which represents the exchange of longitudinal and scalar photons, corresponds to the interaction e2 VC D 0

Z

i d3 k h ei kr 12 O O : 1 .˛ k/.˛ k/ 1 2 .2/3 q 2 k 2 C i

(F.69)

This Fourier transform is evaluated in Appendix J.3 and yields VC D

e2 0

Z

d3 k .2/3

Z q2 e2 ei kr 12 d3 k ei kr 12 1 2 D k q 2 k 2 C i 0 .2/3 k 2 i

(F.70)

provided that the orbitals are generated in a local potential. Using the transform in Appendix J.2, this becomes VCoul D

e2 : 40 r12

(F.71)

Thus, we see that the exchange of longitudinal and scalar photons corresponds to the instantaneous Coulomb interaction, while the exchange of the transverse photons corresponds to the Breit interaction. Note that this is true only if the orbitals are generated in a local potential.

324

F Semiclassical Theory of Radiation

If instead of the separation (F.68), we would separate the photons into the scalar part .p D 0/ and the vector part .p D 1; 2; 3/,

1 ˛1 ˛2 D

8 t2 > t1 > 1; which corresponds to the evolution operator (I.1) Z 2

t

it2 d2

.i/

1

dt2 V .t2 / e

Z

t2 1

dt1 V .t1 / eit1 d1 :

(I.3)

The contraction of the radiation-field operators gives rise to a photon propagator (4.18) A .x2 /A .x1 / D i DF .x2 ; x1 / and this leads to the interaction (4.46) I.x2 ; x1 / D e 2 c 2 ˛1 ˛2 DF .x2 ; x1 / D

Z

dz iz.t2 t1 / e 2

Z

2c 2 k dk f .kI x 1 ; x 2 / : z2 c 2 k 2 C i

The time dependence at vertex 1 then becomes eit1 d1 , where d1 D "a "r z C i: This parameter is referred to as the vertex value and given by the incoming minus the outgoing orbital energies/energy parameters at the vertex. Similarly, we define d2 D "b "s C z C i d12 D d1 C d2 D E "r "s

(I.4)

with E D "a C "b . This leads to the time integrals Z .i/

2

t 1

dt2 e

it2 d2

Z

t2 1

dt1 eit1 d1 D

eit d12 1 : d12 d1

(I.5)

I.2 Two-Photon Exchange

343

Together with the opposite time ordering, t > t1 > t2 > 1, the denominators become 1 1 1 1 1 1 C : (I.6) D C d12 d1 d2 E "r "s "a "r z C i "b "s C z C i This leads to the Feynman amplitude (6.11) dz 1 1 1 C Msp D i 2 E "r "s "a "r z C i "b "s C z C i Z 2c 2 k dk f .k/ (I.7) z2 c 2 k 2 C i Z

or hrsjMspjabi D

1 hrsjVsp .E/jabi E "r "s

(I.8)

with

Z Vsp .E/ D

c dk f .k/

1 1 C "a "r ck C i "b "s ck C i

:

(I.9)

If the interaction is instantaneous, then the time integral becomes Z i

t 1

dt12 eit12 d12 D

eit d12 eit .E"r "s / D ; d12 E " r "s

(I.10)

which for t D 0 is the standard MBPT result.

I.2 Two-Photon Exchange Next, we consider the diagrams in Fig. I.1. We extend the definitions of the vertex values: d1 D " a " t z I

d 2 D "b " u C z I

d12 D d1 C d2 D E "t "u I d123 D E "r "u z0 I

d3 D "t "r z0

d13 D "a "r z z0 I d124 D E "t "s C z0 I

d4 D "u "s C z0 ;

d24 D "b "s C z C z0 ; d1234 D E "r "s ;

i.e., given by the incoming minus the outgoing energies of the vertex. There is a damping term ˙i for integration going to 1.

344

t

I Evaluation Rules for Time-Ordered Diagrams

r

s

r

s 4

0

z -

r

1 a

1

s

-z

t

u

r

s z -

r

E D "a C "b

t

4

1 t

r

s

r

s z

u

3

2 b

a

b

t

4

3

2

z -

s

r

u

3 t

t

2

u

1

E

0

t

2 b

a

a

E

z 3

4 b

E

Fig. I.1 Time-ordered Feynman diagrams for the two-photon ladder with only particle states (left) and with one and two intermediate hole states (right)

I.2.1 No Virtual Pair We consider now the first diagram above, where only particle states are involved. We assume that it is reducible, implying that the two photons do not overlap in time. Then the time ordering is t > t4 > t3 > t2 > t1 > 1: This leads to the time integrations Z .i/4

t 1

dt4 eit4 d4

Z

t4

1

eit d1234 D : d1234 d123 d12 d1

dt3 eit3 d3

Z

t3 1

dt2 eit2 d2

Z

t2 1

dt1 eit1 d1 (I.11)

Changing the order between t1 and t2 and between t3 and t4 leads to the denominators 1 1 1 1 1 1 C C : (I.12) d1234 d123 d124 d12 d1 d2 Here, all integrations are being performed upward, which implies that all denominators are evaluated from below. If the interaction 1–2 is instantaneous, then the integrations become Z t Z t4 Z t3 eit d1234 3 it4 d4 it3 d3 dt4 e dt3 e dt12 eit12 d12 D .i/ d1234 d123 d12 1 1 1 (I.13)

I.2 Two-Photon Exchange

345

and together with the other time ordering eit d1234 1 1 1 C : d1234 d123 d124 d12

(I.14)

If both interactions are instantaneous, we have Z

t

.i/2

1

Z

dt34 eit34 d34

t34

dt12 eit12 d12 D

1

eit d1234 d1234 d12

(I.15)

consistent with the MBPT result [1, Sect. 12.2].

I.2.2 Single Hole Next, we consider the two-photon exchange with a single hole, represented by the second diagram above. We still assume that the diagram is reducible, implying that the two photons do not overlap in time. The time ordering is now t > t4 > t3 > t2 > 1

1 > t1 > t4 ;

and

but the order between t1 and t is not given. If this is considered as a Goldstone diagram, all times (including t1 ) are restricted to tn < t, which leads to Z

1

dt1 e

t

D

it1 d1

Z

t1

it4 d4

1

Z

dt4 e

t4

it3 d3

1

Z

dt3 e

t3 1

dt2 eit2 d2

eit d1234 : d1234 d234 d23 d2

(I.16)

Note that the last integration is being performed in the negative direction, due to the core hole. This is illustrated in Fig. I.2 (left). Considered as a Feynman diagram, the time t1 can run to C1, which leads to Z

t4

1

D

it1 d1

dt1 e

Z

t

1

dt4 e

eit d1234 : d1234 d1 d23 d2

it4 d4

Z

t4

1

it3 d3

dt3 e

Z

t3

1

dt2 eit2 d2 (I.17)

Here, the last integration is still performed in the negative direction, this time from C1 to t4 , and this leads to a result different from the previous one. In the Goldstone case all denominators are evaluated from below, while in the Feynman case one of them is evaluated from above (see Fig. I.2, right). For diagrams diagonal in energy we have d1234 D 0, and hence d1 D d234 , which implies that in this case the two results are identical.

346

I Evaluation Rules for Time-Ordered Diagrams t

t r

s

1

?

d234

-z

t

r

d1234

4 d23

3

?

d2

s

1 t

?

d1234

6

4 d23

3

d2

?

? ?

2 b

a

?

d1

-z

2 b

a

Feynman

Goldstone

Fig. I.2 Time-ordered Goldstone and Feynman diagrams, respectively, for two-photon exchange with one virtual pair. In the latter case, one denominator (at vertex 1) is evaluated from above

Let us next consider the third diagram in Fig. I.1, where the time ordering is t > t4 > t1 > t3 > t2 > 1: Here, all times are limited from above in the Goldstone as well as the Feynman interpretation, and this leads in both cases to Z

t

it4 d4

1

dt4 e

D

Z

1

it1 d1

Z

dt1 e

t4

t1

it3 d3

1

Z

dt3 e

t3 1

dt2 eit2 d2

eit d1234 : d1234 d123 d23 d2

(I.18)

I.2.3 Double Holes The last diagram in Fig. I.1, also reproduced in Fig. I.3, represents double virtual pair. Considered as a Goldstone diagram, the time ordering is t > t2 > t1 > t4 > t3 > 1; which yields Z

t 1

D

it2 d2

dt2 e

Z

t2 1

dt1 e

eit d1234 : d1234 d134 d34 d3

it1 d1

Z

t1

1

it4 d4

dt4 e

Z

t4 1

dt3 eit3 d3 (I.19)

I.3 General Evaluation Rules

347

t

t r

s

d1234 2

z

d134

r z

z0 -

4

d3

3

a

b

? ?

d12 z0 -

t a

d2

?

6

u

1

d34 t

d1234 2

?

u

1

s

?

4

3

d3 b

6 ?

Feynman

Goldstone

Fig. I.3 Time-ordered Goldstone and Feynman diagrams, respectively, for two-photon exchange with two virtual pairs. In the latter case, two denominators (at vertices 1 and 2) are evaluated from above

This is illustrated in Fig. I.3 (left). Considered as a Feynman diagram, we have instead 1 > t2 > t1 , which leads to Z

t 1

D

it4 d4

dt4 e

Z

t4

1

dt3 e

it3 d3

Z

t4 1

it1 d1

dt1 e

Z

t1 1

eit d1234 ; d1234 d3 d12 d2

dt2 eit2 d2 (I.20)

where two integrations are performed in the negative direction.

I.3 General Evaluation Rules We can now formulate evaluation rules for the two types of diagrams considered here. For (nonrelativistic) Goldstone diagrams, the rules are equivalent to the standard Goldstone rules [1, Sect. 12.2] There is a matrix element for each interaction. For each vertex, there is a denominator equal to the vertex sum (sum of vertex

values): incoming minus outgoing orbital energies and z C i for crossing photon line (leading to ck after integration) below a line immediately above the vertex. For particle/hole lines, the integration is performed in the positive/negative direction.

348

I Evaluation Rules for Time-Ordered Diagrams

For the relativistic Feynman diagrams, the same rules hold, with the exception that For a vertex where time can run to C1, the denominator should be evaluated

from above with the denominator equal to the vertex sum above a line immediately below the vertex (with z i for crossing photon line, leading to Cck).

References 1. Lindgren, I., Morrison, J.: Atomic Many-Body Theory. Second edition, Springer-Verlag, Berlin (1986, reprinted 2009) ˚ en, B.: The covariant-evolution-operator method in bound2. Lindgren, I., Salomonson, S., As´ state QED. Physics Reports 389, 161–261 (2004)

Appendix J

Some Integrals

J.1 Feynman Integrals In this section, we shall derive some integrals, which simplify many QED calculations considerably (see the books of Mandl and Shaw [1, Chap. 10] and Sakurai [2, Appendix D]), and we shall start by deriving some formulas due to Feynman. We start with the identity 1 1 D ab ba

Z

b

a

dt : t2

(J.1)

With the substitution t D b C .a b/x, this becomes 1 D ab

Z

1 0

dx D Œb C .a b/x2

Z 0

1

dx : Œa C .b a/x2

(J.2)

Differentiation with respect to a yields 1 D2 a2 b

Z

1

0

xdx : Œb C .a b/x3

(J.3)

Similarly, we have Z

Z

1

x

1 Œa C .b a/x C .c b/y3 0 0 Z d Z 1x 1 D2 x dy : Œa C .b a/x C .c a/y3 0 0

1 D2 abc

dx

dy

(J.4)

Next we consider the integral Z

1 d k 2 D 4 .k C s C i /3 4

Z

Z

1

jkj djkj 2

1

dk0 : .k 2 C s C i /3

349

350

J Some Integrals

The second integral can be evaluated by starting with Z 1 i dk0 Dp 2 2 jkj2 s 1 k0 jkj C s C i evaluated by residue calculus, and differentiating twice with respect to s. The integral then becomes Z Z jkj2 djkj 1 3i 2 i 2 4 : (J.5) d k 2 D D .k C s C i /3 2 2s .jkj2 C s/5=2 The second integral can be evaluated from the identity x2 1 s D 2 2 2 5=2 3=2 .x C s/ .x C s/ .x C s/5=2 and differentiating the integral Z p dx D ln x C x 2 C s p x2 C s yielding

Z

1 x2 D : 3s .x 2 C s/5=2

For symmetry reason, we find Z d4 k

k D 0: .k 2 C s C i /3

(J.6)

Differentiating this relation with respect to k leads to Z d4 k

kk g D .k 2 C s C i /4 3

Z d4 k

1 i 2 g D .k 2 C s C i /3 6s

(J.7)

using the relation (A.4). By making the replacements k )kCq

qnd s ) s q 2 ;

the integrals (J.5) and (J.6) lead to Z 1 i 2 D d4 k 2 ; .k C 2kq C s C i /3 2.s q 2 / Z

(J.8)

Z k q i 2 q 4 d k 2 : D d k D .k C 2kq C s C i /3 .k 2 C 2kq C s C i /3 2.s q 2 / (J.9) 4

J.2 Evaluation of the Integral

R

d 3k eikr12 .2/3 q 2 k 2 Ci

351

Differentiating the last relation with respect to q leads to Z kk i 2 2q q g D C ‹‹: d4 k 2 .k C 2kq C s C i /4 12 s q 2 .s q 2 /2 Differentiating the relation (J.8) with respect to s yields Z 1 i 2 d4 k 2 D ; 4 .k C 2kq C s C i / 6.s q 2 /2 which can be generalized to arbitrary integer powers 3 Z 1 1 .n 3/Š D i 2 : d4 k 2 n .k C 2kq C s C i / .n 1/Š .s q 2 /n2 This can also be extended to nonintegral powers Z 1 .n 2/ 1 d4 k 2 D i 2 .k C 2kq C s C i /n .n/ .s q 2 /n2

(J.10)

(J.11)

(J.12)

(J.13)

and similarly Z d4 k Z

.k 2

k q .n 2/ D i 2 ; n C 2kq C s C i / .n/ .s q 2 /n2

(J.14)

kk g 2 .n 3/ .2n 3/ q q: D i C d k 2 ; .k C 2kq C s C i /n 2 .n/ .s q 2 /n2 .s q 2 /n3 (J.15) 4

Z J.2 Evaluation of the Integral

d 3k eikr12 .2/3 q 2 k2 C i

Using spherical coordinates k D . ; ; /, . D jkj/, we have with d3 k D 2 d sin d d˚ and r12 D jx 1 x 2 j Z

Z

Z

2 d d sin ei r12 cos 2 2 C i q 0 0

Z 1

d ei r12 ei r12 i D 2 4 r12 0 q 2 2 C i

Z 1

d ei r12 ei r12 i D 2 ; (J.16) 8 r12 1 q 2 2 C i

d3 k eik.x1 x 2 / D .2/2 .2/3 q 2 k2 C i

1

352

J Some Integrals

where we have in the last step used the fact that the integrand is an eve n function of . The poles appear at D ˙q.1 C i =2q/. ei r12 is integrated over the posi ti ve and ei r12 over the negati ve half-plane, which yields e˙iqr12 =.4 r12 / with the upper sign for q > 0. The same result is obtained if we change the sign of the exponent in the numerator of the original integrand. Thus, we have the result Z

1 d3 k e˙ik.x1 x2 / D .2/3 q 2 k2 C i 4 2 r12

Z

1

0

eijqjr12 2 d sin. r12 / : D q 2 2 C i 4 r12

(J.17)

The imaginary part of the integrand, which is an odd function, does not contribute to the integral.

J.3 Evaluation of the Integral Z eikr12 d 3k O O ˛ k ˛ k 1 2 .2/3 q 2 k2 C i The integral appearing in the derivation of the Breit interaction (F.53) is Z

eikr 12 d3 k .˛ k/.˛ k/ 1 2 .2/3 q 2 k2 C i Z eikr 12 d3 k D .˛1 r 1 /.˛2 r 2 / : 2 3 .2/ k .q 2 k2 C i /

I2 D

(J.18)

Using (J.16), we then have I2 D D

i 8 2 r12

Z .˛1 r 1 /.˛2 r 2 /

1 .˛1 r 1 /.˛2 r 2 / 4 2 r12

Z

1 1

1 0

d ei r12 ei r12 .q 2 2 C i /

2 d sin.kr12 / :

2 .q 2 2 C i /

(J.19)

The poles appear at D 0 and D ˙.q C i =2q/. The pole at D 0 can be treated with half the pole value in each half plane. For q > 0, the result becomes

1 eiqr12 1 4 r12 q 2

and for q > 0 the same result with q in the exponent. The final result then becomes

J.3 Evaluation of the Integral

Z

R

d 3k .2/3

˛1 kO

˛2 kO

eikr12 q 2 k 2 Ci

353

1 d3 k eikr 12 eijqjr12 1 D .˛ k/.˛ k/ .˛ r /.˛ r / 1 2 1 1 2 2 .2/3 4 r12 q2 q 2 k2 C i D

1 .˛1 r 1 /.˛2 r 2 / 4 2 r12 Z 1 2 d sin. r12 / :

2 .q 2 2 C i / 0

(J.20)

Assuming that our basis functions are eigenfunctions of the Dirac hamiltonian hO D , we can process this integral further. Then the commutator with an arbitrary function of the space coordinates is i h hO D ; f .x/ D c ˛ pO f .x/ C ŒU; f .x/ :

(J.21)

The last term vanishes if the potential U is a local function, yielding h

i hO D ; f .x/ D c ˛ pO f .x/ D ic ˛ r f .x/ :

(J.22)

In particular h

i hO D ; eikx D ic ˛ r eikx D c ˛ kO eikx :

(J.23)

We then find that .˛ r /1 .˛ r /2 eikx D

1 hD ; eikx 1 hD ; eikx 2 2 c

(J.24)

with the matrix element hrsj .˛ r /1 .˛ r /2 eikx jabi D q 2 eikx

(J.25)

using the notation in (F.49). The integral (J.18) then becomes Z I2 D

eikr 12 d3 k O O D .˛ k/.˛ k/ 1 2 .2/3 q 2 k2 C i

Z

eikr 12 d3 k q 2 .2/3 k2 q 2 k2 C i (J.26)

provided that the orbitals are generated by a hamiltonian with a local potential.

354

J Some Integrals

References 1. Mandl, F., Shaw, G.: Quantum Field Theory. John Wiley and Sons, New York (1986) 2. Sakurai, J.J.: Advanced Quantum Mechanics. Addison-Wesley Publ. Co., Reading, Mass. (1967)

Appendix K

Unit Systems and Dimensional Analysis

K.1 Unit Systems K.1.1 SI System The standard unit system internationally agreed upon is the SI system or System Internationale.1 The basis units in this system are given in the following table:

Quantity Length Mass Time Electric current Thermodynamic temperature Amount of substance Luminous intensity

SI unit Meter Kilogram Second Ampere Kelvin Mole Candela

Symbol m kg s A K mol cd

For the definition of these units, the reader is referred to the NIST WEB page (see footnote). From the basis units – particularly the first four – the units for most other physical quantities can be derived.

K.1.2 Relativistic or “Natural” Unit System In scientific literature, some simplified unit system is frequently used for convenience. In relativistic field theory the relativistic unit system is mostly used, where the first four units of the SI system are replaced by

1

For further details, see The NIST Reference on Constants, Units, and Uncertainty (http://physics.nist.gov/cuu/Units/index.html). 355

356

K Unit Systems and Dimensional Analysis Quantity Mass Velocity Action Dielectricity

Relativistic unit Rest mass of the electron Light velocity in vacuum Planck’s constant divided by 2 Dielectricity constant of vacuum

Symbol m c „ 0

Dimension kg ms1 kg m2 s1 A2 s4 kg1 m3

In the table, the dimension of the relativistic units in SI units are also shown. From these four units, all units that depend only on the four SI units kg, s, m, A can be derived. For instance, energy that has the dimension kg m2 m2 has the relativistic unit me c 2 , which is the rest energy of the electron ( 511 keV). The unit for length is „ D =2 0; 386 1012 m; me c where is Compton wavelength and the unit for time is 2c= 7; 77 104 s).

K.1.3 Hartree Atomic Unit System In atomic physics, the Hartree atomic unit system is frequently used, based on the following four units Quantity Mass Electric charge Action Dielectricity

Atomic unit Rest mass of the electron Absolute charge of the electron Planck’s constant divided by 2 Dielectricity constant of vacuum times 4

Symbol m e „ 40

Dimension kg As kg m2 s1 A2 s4 kg1 m3

Here, the unit for energy becomes 1H D

me 4 ; .40 /2 „3

which is known as the Hartree unit and equals twice the ionization energy of the hydrogen atom in its ground state ( 27.2 eV). The atomic unit for length is a0 D

40 „2 me 2

known as the Bohr radius or the radius of the first electron orbit of the Bohr hydrogen model ( 0:529 1010 m). The atomic unit of velocity is ˛c, where ˛D

e2 40 c„

(K.1)

K.2 Dimensional Analysis

357

is the dimensionless fine-structure constant ( 1/137,036). Many units in these two systems are related by the fine-structure constant. For instance, the relativistic length unit is ˛a0 .

K.1.4 cgs Unit Systems In older scientific literature, a unit system, known as the cgs system, was frequently used. This is based on the following three units: Quantity Length Mass Time

cgs Unit Centimeter Gram Second

Symbol cm g s

In addition to the three units, it is necessary to define a fourth unit to be able to derive most of the physical units. Here, two conventions are used. In the electrostatic version (ecgs), the proportionality constant of Coulombs law, 40 , is set equal to unity, and in the magnetic version (mcgs) the corresponding magnetic constant, 0 =4, equals unity. Since these constants have dimension, the systems cannot be used for dimensional analysis (see below). The most frequently used unit system of cgs type is the so-called Gaussian unit system, where electric units are measured in ecgs and magnetic ones in mcgs. This implies that certain formulas will look differently in this system, compared to a system with consistent units. For instance, the Bohr magneton, which in any consistent unit system will have the expression B D

e„ 2m

will in the mixed Gaussian system have the expression B D

e„ 2mc

which does not have the correct dimension. Obviously, such a unit system can easily lead to misunderstandings and should be avoided.

K.2 Dimensional Analysis It is often useful to check physical formulas by means of dimensional analysis, which, of course, requires that a consistent unit system, like the SI system, is being

358

K Unit Systems and Dimensional Analysis

used. Below we list a number of physical quantities and their dimension, expressed in SI units, which could be helpful in performing such an analysis. In most parts of the book, we have set „ D 1, which simplifies the formulas. This also simplifies the dimensional analysis, and in the last column below we have (after the sign )) listed the dimensions in that case.

Œforce D N D

kg m 1 ; ) 2 s ms

Œenergy D J D Nm D Œaction; „ D Js D

kg m2 1 ) ; s2 s

kg m2 ) 1; s

J kg m2 1 D ) ; As As3 As2 1 kg m ) ; Œelectric field; E D V=m D As3 Ams2 kg 1 Œmagnetic field; B D Vs=m2 D ; ) 2 As Am2 s kg m 1 Œvector potential; A D Vs=m D ) ; 2 Ams As 1 kg m ) ; Œmomentum; p D s m As Œcharge density; D 3 ; m A Œcurrent density; j D 2 ; m kg m 1 Œ0 D N=A2 D 2 2 ) 2 ; A s A ms

Œelectric potential D V D

Œ0 D Œ1=0 c 2 D

A2 s4 3

kg m

)

A2 s 3 : m

Fourier transforms Z DF .x1 ; x2 / D

dz DF .zI x 1 ; x 2 / eiz.t2 t1 / ; 2

ŒA.!; x/ D sŒA.x/; ŒA.!; k/ D sm3 ŒA.x/; ŒA.k/ D m4 ŒA.x/:

K.2 Dimensional Analysis

359

Photon propagator ŒDF .x; x/ D

1 A2 m 2 s 2

;

s ; m3 Œ0 DF .k/ D sm; s Œ0 DF .k0 ; x/ D 2 ; m 1 Œ0 DF .t; x/ D 2 ; m Œ0 DF .x; x/ D

Œ0 DF .z; x/ D

s2 m3

z D ck0 ;

Œ0 DF .z; k/ D s2 ; Œe 2 c 2 DF .z; x/ D I.z; x/ D

m e2 D : 0 s

Electron propagator SOF .x; x/ ) s; 1 ; m3 s SF .z; x/ ) 3 ; m SF .z; k/ ) s;

SF .x; x/ )

SF .k/ ) m:

S-matrix SO .x; x/ ) 1; S.z; x/ ) s; S.z; k/ ) m4 :

1 ; s

360

K Unit Systems and Dimensional Analysis

Self energy 1 ˙O .z/ ) ; s 1 ˙.z; x/ ) ; s m3 1 ˙.z; k/ ) ; s m ˙.k/ ) 2 : s Vertex .z; k/ ) 1; m .p; p 0 / ) : s

Abbreviations

CCA

Coupled-cluster approach

CEO

Covariant evolution operator

GML

Gell-Mann–Low relation

HP

Heisenberg picture

IP

Interaction picture

LDE

Linked-diagram expansion

MBPT

Many-body perturbation theory

MSC

Model-space contribution

NVPA

No-virtual-pair approximation

PWR

Partial-wave regularization

QED

Quantum electrodynamics

SCF

Self-consistent field

SI

International unit system

SP

Schrödinger picture

361

Index

A Adiabatic damping, 51 all-order method, 30 annihilation operator, 283 antisymmetry, 22

covariant evolution operator, 3, 119 covariant vector, 273 creation operator, 283 cut-off procedure, 248

B Banach space, 277 Bethe–Salpeter equation, 3, 119, 193, 199, 225 effective potential form, 205 Bethe–Salpeter–Bloch equation, 6, 156, 173, 193, 199, 206 Bloch equation for Green’s operator, 146 generalized, 20, 23 bra vector, 290 Breit interaction, 38, 73, 318 Brillouin–Wigner expansion, 153 Brown–Ravenhall effect, 2, 38, 178, 194, 216, 219

D d’Alembertian operator, 274 de Broglie’s relations, 13 density operator, 120 difference ratio, 139 dimensional analysis, 355 dimensional regularization, 163 Dirac delta function, 277 Dirac equation, 1, 295 Dirac matrices, 296 Dirac sea, 38 Dirac-Coulomb Hamiltonian, 37 discretization technique, 42 dot product, 135 Dyson equation, 108, 201, 243

C Cauchy sequence, 276 closure property, 292 complex rotation, 37 configuration, 22 conjugate momentum, 15, 306 connectivity, 37 continuity equation, 314 contraction, 18 contravariant vector, 273 coordinate representation, 63, 292 counterterms, 138 coupled-cluster approach, 30 normal-ordered, 34 coupled-cluster-QED expansion, 196 covariance, 1

E effective Hamiltonian, 5, 20 intermediate, 37 effective interaction, 23, 31 Einstein summation rule, 289 electron field operators, 285 electron propagator, 61, 98 equal-time approximation, 96, 120, 124, 173, 205, 208 Euler-Lagrange equations, 308 evolution operator, 47 exponential Ansatz, 33 normal-ordered, 34 external-potential approach, 4, 208

363

364 F Feynman amplitude, 88, 96, 124, 174, 335 Feynman diagram, 28, 60, 87, 125, 335 fine structure, 165 first quantization, 13 Fock space, 133, 277 photonic, 6, 51, 132, 211 fold, 126 folded diagram, 23 Fourier transform, 63 functional, 275 Furry picture, 21, 27 Furry’s theorem, 86

G g-factor, 164 gamma function, 332 gauge Coulomb, 59, 68, 76, 81, 315, 324, 329 covariant, 59, 65, 122, 125, 328 Feynman, 59, 65, 66, 81, 324, 326, 328 Fried-Yennie, 329 Landau, 328 non-covariant, 329 gauge invariance, 314 gauge transformation, 327 Gaunt interaction, 213, 217 Gell-Mann–Low theorem, 51 relativistic, 131 Goldstone diagram, 25 Goldstone rules, 24, 28, 125 Green’s function, 3, 91 projected, 113 Green’s operator, 119, 134 Grotsch term, 164 Gupta-Bleuler formalism, 325

H Hamiltonian density, 309 Hartree–Fock model, 26 Heaviside step function, 282 Heisenberg picture, 202, 286 Heisenberg representation, 92 helium fine structure, 225 Hilbert space, 277 hole state, 119 Hylleraas function, 5, 163

I instantaneous approximation, 4 interaction picture, 47

Index intermediate normalization, 21 intruder state, 23, 36 irreducible diagram, 40, 127

K ket vector, 290 Klein-Gordon equation, 295 Kronecker delta factor, 278

L Lagrange equations, 305 Lagrangian function, 309 Lamb shift, 2, 79, 84, 157 Laplacian operator, 274 Lehmann representation, 97 linked diagram, 28, 107 Lippmann–Schwinger equation, 205 Lorentz covariance, 1, 39, 60, 94, 119 Lorentz force, 308 Lorentz transformation, 1 Lorenz condition, 314

M many-body Dirac Hamiltonian, 133 matrix elements, 291 matrix representation, 291 Maxwell’s equations, 311, 312 metric tensor, 273 minimal substitutions, 307 model space, 20 complete, 22 extended, 23 model state, 20 model-space contribution, 29, 56, 126, 138, 142 momentum representation, 300 MSC, 56, 126 multi-photon exchange, 127

N no-virtual-pair approximation, 2, 37 non-radiative effects, 39 norm, 275 normal order, 18

P pair correlation, 30 parent state, 52, 131 partitioning, 21

Index Pauli spin matrices, 296 perturbation Brillouin–Wigner, 111 Rayleigh–Schrödinger, 1, 24 photon propagator, 65 Poisson bracket, 15, 306 polarization tensor, 87 principal-value integration, 64

Q QED effects, 39, 59, 79 QED potential, 181 quantization condition, 15 quasi-degeneracy, 23, 56 quasi-potential approximation, 4 quasi-singularity, 56

R radiative effects, 39 reducible diagram, 40, 127 reference-state contribution, 78, 142 regularization, 79, 237 Brown–Langer–Schaefer, 252 dimensional, 257 partial-wave, 255 Pauli–Willars, 248 renormalization, 79, 237 charge, 244 mass, 242 resolution of the identity, 290 resolvent, 24 reduced, 24

S S-matrix, 3, 60, 157 scalar potential, 311 scalar product, 274 scalar retardation, 213, 217 scattering matrix, 60 Schrödinger equation, 15 Schrödinger picture, 47 Schwinger correction, 164 second quantization, 16, 283 self energy electron, 39 self energy, 108 electron, 79, 157, 186, 239, 244

365 photon, 87, 246 proper, 108 sequence, 276 set, 275 size consistency, 30 size extensive, 207 size extensivity, 30, 207 Slater determinant, 22, 284 spin-orbital, 22 spline, 41 state-specific approach, 37 state-universality, 36 subset, 275 Sucher energy formula, 61

T target states, 20 time-ordering, 49 Wick, 18 trace, 85 transverse-photon, 77 two-times Green’s function, 3, 111

U Uehling potential, 86, 128 union, 275 unit system, 355 cgs, 357 Hartree, 356 mixed, 300 natural, 355 relativistic, 59, 355 SI , 355 unlinked diagram, 28

V vacuum polarization, 2, 39, 84, 87 valence universality, 21, 36 vector potential, 311 vector space, 275 vertex correction, 39, 82, 186, 240, 246

W Ward identity, 83, 242 wave operator, 20 Wick’s theorem, 19 Wickmann-Kroll potential, 86

Relativistic Many-Body Theory: A New Field-Theoretical Approach

Behavioral Decision Theory: A New Approach

Behavioral decision theory: A new approach

Relativistic Electron Theory

Astrophysics: A New Approach

Statistics: A New Approach

Astrophysics: A new approach

Blackwork: A New Approach

Relativistic Heavy-Particle Collision Theory

Construction Economics: A New Approach

Analytical Dynamics: A New Approach

Composing music: a new approach

Construction Economics: A New Approach

Composing Music: A New Approach

Location Theory. A Unified Approach

Location Theory: A Unified Approach

Syntactic Theory: A Unified Approach

Organisational Theory: A Practical Approach

Syntactic Theory: A unfied approach

Feynman's Thesis: A New Approach to Quantum Theory

Location theory: a unified approach

Feynman's Thesis: A New Approach to Quantum Theory

Operational Quantum Theory II: Relativistic Structures

Relativistic quantum mechanics and field theory

The Relativistic Boltzmann Equation: Theory and Applications

Operational quantum theory I: Non-relativistic structures

Introduction to Relativistic Quantum Field Theory

Relativistic Quantum Mechanics and Field Theory

Introduction to Relativistic Quantum Field Theory

Operational quantum theory II: Relativistic structures

Introduction to Relativistic Quantum Field Theory

Relativistic Many-Body Theory: A New Field-Theoretical Approach

Behavioral Decision Theory: A New Approach

Behavioral decision theory: A new approach

Relativistic Electron Theory

Astrophysics: A New Approach

Statistics: A New Approach

Astrophysics: A new approach

Blackwork: A New Approach

Relativistic Heavy-Particle Collision Theory

Construction Economics: A New Approach

Analytical Dynamics: A New Approach

Composing music: a new approach

Construction Economics: A New Approach

Composing Music: A New Approach

Location Theory. A Unified Approach

Location Theory: A Unified Approach

Syntactic Theory: A Unified Approach

Organisational Theory: A Practical Approach

Syntactic Theory: A unfied approach

Feynman's Thesis: A New Approach to Quantum Theory

Location theory: a unified approach

Feynman's Thesis: A New Approach to Quantum Theory

Operational Quantum Theory II: Relativistic Structures

Relativistic quantum mechanics and field theory

The Relativistic Boltzmann Equation: Theory and Applications

Operational quantum theory I: Non-relativistic structures

Introduction to Relativistic Quantum Field Theory

Relativistic Quantum Mechanics and Field Theory

Introduction to Relativistic Quantum Field Theory

Operational quantum theory II: Relativistic structures

Introduction to Relativistic Quantum Field Theory

Recommend Documents